Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks502 from public data
- PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud3,019
Extensive experiments on the 3D detection benchmark of KITTI dataset show that the proposed architecture outperforms state-of-the-art methods with remarkable margins by using only point cloud as input.
- StackGAN: Text to Photo-Realistic Image Synthesis with Stacked Generative Adversarial Networks2,975
This paper proposes Stacked Generative Adversarial Networks (StackGAN) to generate 256 photo-realistic images conditioned on text descriptions and introduces a novel Conditioning Augmentation technique that encourages smoothness in the latent conditioning manifold.
- PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection2,374
The proposed PV-RCNN surpasses state-of-the-art 3D detection methods with remarkable margins and deeply integrates both 3D voxel Convolutional Neural Network and PointNet-based set abstraction to learn more discriminative point cloud features.
- CLIP-Adapter: Better Vision-Language Models with Feature Adapters1,954
This paper proposes CLIP-Adapter, a novel approach to vision-language pretraining that adopts an additional bottleneck layer to learn new features and performs residual-style feature blending with the original pretrained features and is able to outperform context optimization while maintaining a simple design.
- InternImage: Exploring Large-Scale Vision Foundation Models with Deformable Convolutions1,217
This work presents a new large-scale CNN-based foundation model, termed InternImage, which can obtain the gain from increasing parameters and training data like ViTs, and reduces the strict inductive bias of traditional CNNs and makes it possible to learn stronger and more robust patterns with large- scale parameters from massive datalike ViTs.
- StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks1,198
Extensive experiments demonstrate that the proposed stacked generative adversarial networks significantly outperform other state-of-the-art methods in generating photo-realistic images.
- 1,193
- LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention1,068
A zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge on traditional vision and language tasks, demonstrating the superior generalization capacity of the approach.
- From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network1,046
This paper extends the preliminary work PointRCNN to a novel and strong point-cloud-based 3D object detection framework, the part-aware and aggregation neural network, which outperforms all existing 3D detection methods and achieves new state-of-the-art on KITTI 3D objects detection dataset by utilizing only the LiDAR point cloud data.
- Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identification997
This work presents a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks with CNNs and proposes a Domain Guided Dropout algorithm to improve the feature learning procedure.
- 995
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis978
The Human Preference Score v2 (HPS v2) is obtained, a scoring model that can more accurately predict human preferences on generated images and is responsive to algorithmic improvements of text-to-image generative models, making it a preferable evaluation metric for these models.
- Balanced Meta-Softmax for Long-Tailed Visual Recognition868
Balanced Softmax is presented, an elegant unbiased extension of Softmax, to accommodate the label distribution shift between training and testing, and it is demonstrated that Balanced Meta-Softmax outperforms state-of-the-art long-tailed classification solutions on both visual recognition and instance segmentation tasks.
- Efficient Attention: Attention with Linear Complexities822
A novel efficient attention mechanism equivalent to dot-product attention but with substantially less memory and computational costs is proposed, which allows more widespread and flexible integration of attention modules into a network, which leads to better accuracies.
- LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model774
This work augments LLaMA-Adapter by unlocking more learnable parameters and proposes an early fusion strategy to feed visual tokens only into the early LLM layers, contributing to better visual knowledge incorporation and achieves strong multi-modal reasoning with only a small-scale image-text and instruction dataset.
- Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentation756
A new framework for the outdoor LiDAR segmentation is proposed, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern while maintaining these inherent properties.
- Group-Wise Correlation Stereo Network747
Group-wise correlation provides efficient representations for measuring feature similarities and will not lose too much information like full correlation, and preserves better performance when reducing parameters compared with previous methods.
- 725
- Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID691
This work proposes a novel self-paced contrastive learning framework with hybrid memory that outperforms state-of-the-arts on multiple domain adaptation tasks of object re-ID and even boosts the performance on the source domain without any extra annotations.
- UniFormer: Unifying Convolution and Self-Attention for Visual Recognition646
This work proposes a novel Unified transFormer (UniFormer), which can seamlessly integrate the merits of convolution and self-attention in a concise transformer format, and builds an efficient UniFormer with a concise hourglass design of token shrinking and recovering, which achieves high throughput.
- PV-RCNN++: Point-Voxel Feature Set Abstraction With Local Vector Representation for 3D Object Detection631
The experiments demonstrate that the proposed PV-RCNN++ framework achieves state-of-the-art 3D detection performance on the large-scale and highly-competitive Waymo Open Dataset with 10 FPS inference speed on the detection range of 150 m × 150 m.
- Person Search with Natural Language Description593
An Recurrent Neural Network with Gated Neural Attention mechanism (GNA-RNN) is proposed to establish the state-of-the art performance on person search and a large-scale person description dataset with detailed natural language annotations and person samples from various sources is collected.
- T-CNN: Tubelets With Convolutional Neural Networks for Object Detection From Videos548
A deep learning framework that incorporates temporal and contextual information from tubelets obtained in videos, which dramatically improves the baseline performance of existing still-image detection frameworks when they are applied to videos is proposed, called T-CNN.
- Learning Feature Pyramids for Human Pose Estimation536
This work designs a Pyramid Residual Module (PRMs) to enhance the invariance in scales of DCNNs and provides theoretic derivation to extend the current weight initialization scheme to multi-branch network structures.
- 3D Object Detection for Autonomous Driving: A Comprehensive Survey508
A comprehensive survey of the progress in 3D object detection from the aspects of models and sensory inputs, including LiDAR- based, camera-based, and multi-modal detection approaches, and an in-depth analysis of the potentials and challenges in each category of methods are provided.
- CameraCtrl: Enabling Camera Control for Text-to-Video Generation496
Experimental results demonstrate the effectiveness of CameraCtrl in achieving precise camera control with different video generation models, marking a step forward in the pursuit of dynamic and customized video storytelling from textual and camera pose inputs.
- Bi-directional Cross-Modality Feature Propagation with Separation-and-Aggregation Gate for RGB-D Semantic Segmentation478
A unified and efficient Cross-modality Guided Encoder to not only effectively recalibrate RGB feature responses, but also to distill accurate depth information via multiple stages and aggregate the two recalibrated representations alternatively.
- DeepID-Net: Deformable deep convolutional neural networks for object detection452
The proposed approach improves the mean averaged precision obtained by RCNN, which was the state-of-the-art, from 31% to 50.3% on the ILSVRC2014 detection test set.
- Dynamic Fusion With Intra- and Inter-Modality Attention Flow for Visual Question Answering426
A novel method of dynamically fuse multi-modal features with intra- and inter-modality information flow, which alternatively pass dynamic information between and across the visual and language modalities is proposed, which can robustly capture the high-level interactions between language and vision domains.
- 415
- Fast Convergence of DETR with Spatially Modulated Co-Attention414
This work proposes a simple yet effective scheme for improving the DETR framework, namely Spatially Modulated Co-Attention (SMCA) mechanism, which increases DETR’s convergence speed by replacing the original co-attention mechanism in the decoder while keeping other operations in DETR unchanged.
- 3D Human Pose Estimation in the Wild by Adversarial Learning406
An adversarial learning framework is proposed, which distills the 3D human pose structures learned from the fully annotated dataset to in-the-wild images with only 2D pose annotations and designs a geometric descriptor, which computes the pairwise relative locations and distances between body joints, as a new information source for the discriminator.
- Safety-Enhanced Autonomous Driving Using Interpretable Sensor Fusion Transformer401
A safety-enhanced autonomous driving framework, named Interpretable Sensor Fusion Transformer (InterFuser), to fully process and fuse information from multi-modal multi-view sensors for achieving comprehensive scene understanding and adversarial event detection is proposed.
- Personalize Segment Anything Model with One Shot400
A training-free Personalization approach for SAM, termed as PerSAM, which effectively adapt SAM for private use without any training, and can enhance DreamBooth to personalize Stable Diffusion for text-to-image generation, which discards the background disturbance for better target appearance learning.
- Learning Spatial Regularization with Image-Level Supervisions for Multi-label Image Classification393
Analysis of the learned SRN model demonstrates that it can effectively capture both semantic and spatial relations of labels for improving classification performance, and significantly outperforms state-of-the-arts and has strong generalization capability.
- Online Multi-object Tracking Using CNN-Based Single Object Tracker with Spatial-Temporal Attention Mechanism391
A CNN-based framework for online MOT that utilizes the merits of single object trackers in adapting appearance models and searching for target in the next frame and introduces spatial-temporal attention mechanism (STAM) to handle the drift caused by occlusion and interaction among targets.
- Human Preference Score: Better Aligning Text-to-image Models with Human Preference379
This work proposes a simple yet effective method to adapt Stable Diffusion to better align with human preferences and shows that HPS outperforms CLIP in predicting human choices and has good generalization capability toward images generated from other models.
- 370
- 366
- CAMP: Cross-Modal Adaptive Message Passing for Text-Image Retrieval361
This paper proposes Cross-modal Adaptive Message Passing (CAMP), which adaptively controls the information flow for message passing across modalities and properly handles negative pairs and irrelevant information with an adaptive gating scheme.
- 334
- Person Re-identification with Deep Similarity-Guided Graph Neural Network324
This paper proposes a novel deep learning framework, named Similarity-Guided Graph Neural Network (SGGNN), which learns the edge weights with rich labels of gallery instance pairs directly, which provides relation fusion more precise information.
- SPHINX: The Joint Mixing of Weights, Tasks, and Visual Embeddings for Multi-modal Large Language Models313
SPHINX, a versatile multi-modal large language model (MLLM) with a joint mixing of model weights, tuning tasks, and visual embeddings is presented, and an efficient strategy aiming to better capture fine-grained appearances of high-resolution images is proposed.
- Learning Deep Neural Networks for Vehicle Re-ID with Visual-spatio-Temporal Path Proposals303
A two-stage framework that incorporates complex spatio-temporal information for effectively regularizing the re-identification results is proposed and extensive experiments and analysis show the effectiveness of the proposed method and individual components.
- Zoom-in-Net: Deep Mining Lesions for Diabetic Retinopathy Detection281
A convolution neural network based algorithm for simultaneously diagnosing diabetic retinopathy and highlighting suspicious regions and a network termed Zoom-in-Net which mimics the zoom-in process of a clinician to examine the retinal images is proposed.
- Identity-Aware Textual-Visual Matching with Latent Co-attention277
This paper proposes an identity-aware two-stage framework that learns to embed cross-modal features with a novel Cross-Modal Cross-Entropy (CMCE) loss and refines the matching results with a latent co-attention mechanism.
- Structured Feature Learning for Pose Estimation269
A bi-directional tree structured model is proposed, so that the feature channels at a body joint can well receive information from other joints, and the proposed framework improves feature learning substantially.
- 266
- 265
- 264
- Depth Completion From Sparse LiDAR Data With Depth-Normal Constraints259
A unified CNN framework is proposed that models the geometric constraints between depth and surface normal in a diffusion module and predicts the confidence of sparse LiDAR measurements to mitigate the impact of noise.
- Prompt, Generate, Then Cache: Cascade of Foundation Models Makes Strong Few-Shot Learners255
CaFo, a Cascade of Foundation models that incorporates diverse prior knowledge of various pretraining paradigms for better few-shot learning is proposed that can fully unleash the potential of different pre-training methods and unify them to perform state-of-the-art for few- shot classification.
- ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection255
This work presents a new domain adaptive self-training pipeline, named ST3D, for unsupervised domain adaptation on 3D object detection from point clouds, which achieves state-of-the-art performance on all evaluated datasets and even surpasses fully supervised results on KITTI 3Dobject detection benchmark.
- AdaCos: Adaptively Scaling Cosine Logits for Effectively Learning Deep Face Representations255
A novel cosine-based softmax loss is proposed, AdaCos, which is hyperparameter-free and leverages an adaptive scale parameter to automatically strengthen the training supervisions during the training process and outperforms state-of-the-art softmax losses on all the three datasets.
- Motion-I2V: Consistent and Controllable Image-to-Video Generation with Explicit Motion Modeling251
Qualitative and quantitative comparisons demonstrate the advantages of Motion-I2V over prior approaches in consistent and controllable image-to-video generation.
- 249
- Point-Bind & Point-LLM: Aligning Point Cloud with Multi-modality for 3D Understanding, Generation, and Instruction Following248
Point-LLM is presented, the first 3D large language model (LLM) following 3D multi-modal instructions, which injects the semantics of Point-Bind into pre-trained LLMs, e.g., LLaMA, which requires no 3D instruction data, but exhibits superior 3D and multi- modal question-answering capacity.
- 245
- Interpolated Convolutional Networks for 3D Point Cloud Understanding243
A novel Interpolated Convolution operation, InterpConv, is proposed to tackle the point cloud feature learning and understanding problem, to utilize a set of discrete kernel weights and interpolate point features to neighboring kernel-weight coordinates by an interpolation function for convolution.
- End-to-End Object Detection with Adaptive Clustering Transformer227
A novel variant of transformer named Adaptive Clustering Transformer named ACT has been proposed to reduce the computation cost for high-resolution input and achieves a good balance between accuracy and computation cost (FLOPs).
- CORA: Adapting CLIP for Open-Vocabulary Detection with Region Prompting and Anchor Pre-Matching226
This work proposes CORA, a DETR-style framework that adapts CLIP for Open-vocabulary detection by Region prompting and Anchor pre-matching, which mitigates the whole-to-region distribution gap by prompting the region features of the CLIP-based region classifier.
- Improving Referring Expression Grounding With Cross-Modal Attention-Guided Erasing225
A novel cross-modal attention-guided erasing approach is designed, where the most dominant information from either textual or visual domains is discarded to generate difficult training samples online, and to drive the model to discover complementary textual-visual correspondences.
- Learning Monocular Depth by Distilling Cross-Domain Stereo Networks221
This paper proposes to use the stereo matching network as a proxy to learn depth from synthetic data and use predicted stereo disparity maps for supervising the monocular depth estimation network.
- MathCoder: Seamless Code Integration in LLMs for Enhanced Mathematical Reasoning220
This paper proposes a method to fine-tune open-source language models, enabling them to use code for modeling and deriving math equations and, consequently, enhancing their mathematical reasoning abilities, and yields the MathCoder models, a family of models capable of generating code-based solutions for solving challenging math problems.
- Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification219
The effect of code on enhancing LLMs' reasoning capability by introducing different constraints on the Code Usage Frequency of GPT-4 Code Interpreter is explored, and a novel and effective prompting method, explicit \uline{c}ode-based \ULine{s}elf-\uline {v}erification~(CSV), is proposed to further boost the mathematical reasoning potential of GPN.
- 219
- Learning 3D Representations from 2D Pre-Trained Models via Image-to-Point Masked Autoencoders217
This paper proposes an alternative to obtain superior 3D representations from 2D pre-trained models via Image-to-Point Masked Autoencoders, named as I2P-MAE, which leverages the well learned 2D knowledge to guide 3D masked autoencoding, which reconstructs the masked point tokens with an encoder-decoder architecture.
- MonoDETR: Depth-guided Transformer for Monocular 3D Object Detection216
This paper introduces the first DETR framework for Monocular DEtection with a depth-guided TRansformer, named MonoDETR, and modify the vanilla transformer to be depth-aware and guide the whole detection process by contextual depth cues.
- Meta-Transformer: A Unified Framework for Multimodal Learning214
Meta-Transformer is the first framework to perform unified learning across 12 modalities with unpaired data and indicates a promising future for developing unified multimodal intelligence with transformers.
- Self-supervising Fine-Grained Region Similarities for Large-Scale Image Localization210
This work proposes to self-supervise image-to-region similarities in order to fully explore the potential of difficult positive images alongside their sub-regions and outperforms state-of-the-arts on the standard localization benchmarks and shows excellent generalization capability on multiple image retrieval datasets.
- Single View Stereo Matching201
This paper shows for the first time that the monocular depth estimation problem can be reformulated as two sub-problems, a view synthesis procedure followed by stereo matching, with two intriguing properties, namely i) geometrical constraints can be explicitly imposed during inference; ii) demand on labelled depth data can be greatly alleviated.
- 197
- 191
- 3D Sketch-Aware Semantic Scene Completion via Semi-Supervised Structure Prior187
A new geometry-based strategy to embed depth information with low-resolution voxel representation, which could still be able to encode sufficient geometric information, e.g., room layout, object’s sizes and shapes, to infer the invisible areas of the scene with well structure-preserving details is proposed.
- 186
- Actor-Context-Actor Relation Network for Spatio-Temporal Action Localization184
An Actor-Context-Actor Relation Network (ACAR-Net) is designed which builds upon a novel High-order Relation Reasoning Operator and an Actor- Context Feature Bank to enable indirect relation reasoning for spatio-temporal action localization.
- FuseFormer: Fusing Fine-Grained Information in Transformers for Video Inpainting180
This work proposes FuseFormer, a Transformer model designed for video inpainting via fine-grained feature fusion based on novel Soft Split and Soft Composition operations, which surpasses state-of-the-art methods in both quantitative and qualitative evaluations.
- FlowFormer++: Masked Cost Volume Autoencoding for Pretraining Optical Flow Estimation178
This work proposes Masked Cost Volume Autoencoding (MCVA) to enhance FlowFormer by pretraining the cost-volume encoder with a novel MAE scheme, and introduces a block-sharing masking strategy to prevent masked information leakage, as the cost maps of neighboring source pixels are highly correlated.
- 178
- Instruct2Act: Mapping Multi-modality Instructions to Robotic Actions with Large Language Model177
This paper presents Instruct2Act, a framework that utilizes Large Language Models to map multi-modal instructions to sequential actions for robotic manipulation tasks, employing the LLM model to generate Python programs that constitute a comprehensive perception, planning, and action loop for robotic tasks.
- 177
- ReasonNet: End-to-End Driving with Temporal and Global Reasoning172
This paper presents ReasonNet, a novel end-to-end driving framework that extensively exploits both temporal and global information of the driving scene, and outperforms all prior methods on multiple CARLA benchmarks.
- Pedestrian Behavior Understanding and Prediction with Deep Neural Networks168
A deep neural network (Behavior-CNN) is proposed to model pedestrian behaviors in crowded scenes, which has many applications in surveillance and multiple applications demonstrate the effectiveness of Behavior-CNN on pedestrian behavior modeling.
- Vision-RWKV: Efficient and Scalable Visual Perception with RWKV-Like Architectures166
Vision-RWKV (VRWKV), a model adapted from the RWKV model used in the NLP field with necessary modifications for vision tasks, surpasses ViT's performance in image classification and has significantly faster speeds and lower memory usage processing high-resolution inputs.
- ConvMAE: Masked Convolution Meets Masked Autoencoders164
The ConvMAE framework demonstrates that multi-scale hybrid convolution-transformer can learn more discriminative representations via the mask auto-encoding scheme, and adopts the masked convolution to prevent information leakage in the convolution blocks.
- 163
- NDC-Scene: Boost Monocular 3D Semantic Scene Completion in Normalized Device Coordinates Space163
A novel Normalized Device Coordinates scene completion network (NDC-Scene) is devised that directly extends the 2D feature map to a Normalized device coordinates space, rather than to the world space directly, through progressive restoration of the dimension of depth with deconvolution operations.
- Uni-Perceiver: Pre-training Unified Architecture for Generic Perception for Zero-shot and Few-shot Tasks163
Results show that the pre-trained model without any tuning can achieve reasonable performance even on novel tasks, and the performance can be improved to a level close to state-of-the-art methods by conducting prompt tuning on 1% of downstream task data.
- SPHINX-X: Scaling Data and Parameters for a Family of Multi-modal Large Language Models162
The SPHINX framework is modified by removing redundant visual encoders, bypassing fully-padded sub-images with skip tokens, and simplifying multi-stage training into a one-stage all-in-one paradigm to improve the architecture and training efficiency.
- HMS-Net: Hierarchical Multi-Scale Sparsity-Invariant Network for Sparse Depth Completion160
A sparsity-invariant multi-scale encoder-decoder network (HMS-Net) for handling sparse inputs and sparse feature maps is proposed, and a model withoutRGB guidance ranks 1st among all peer-reviewed methods without using RGB information, and the model with RGB guidance ranks 2nd among all RGB-guided methods.
- Gen-L-Video: Multi-Text to Long Video Generation via Temporal Co-Denoising156
This study introduces a novel paradigm, Gen-L-Video, capable of extending off-the-shelf short video diffusion models for generating and editing videos comprising hundreds of frames with diverse semantic segments without introducing additional training, all while preserving content consistency.
- Refining Pseudo Labels with Clustering Consensus over Generations for Unsupervised Object Re-identification155
This work proposes to properly estimate pseudo label similarities between consecutive training generations with clustering consensus and refine pseudo labels with temporally propagated and ensembled pseudo labels to improve classification with dynamically changing classes over generations.
- Improving Deep Visual Representation for Person Re-identification by Global and Local Image-language Association152
This paper proposes to exploit natural language description as additional training supervisions for effective visual features for person re-identification and achieves state-of-the-art performance without utilizing any auxiliary information during testing and shows better performance than other joint embedding methods for the image-language association.
- Remote sensing image fusion via wavelet transform and sparse representation152
The proposed fusion method can well preserve the spectral information and the spatial detail information of the original images, and could well maintain the spectral characteristics of fusion images with a high spatial resolution.
- Deep Group-Shuffling Random Walk for Person Re-identification150
A novel group-shuffling random walk network for fully utilizing the affinity information between gallery images in both the training and testing processes and which outperforms state-of-the-art methods on the Market-1501, CUHK03, and DukeMTMC datasets by large margins.
- Lumina-T2X: Transforming Text into Any Modality, Resolution, and Duration via Flow-based Large Diffusion Transformers147
This technical report introduces the Lumina-T2X family - a series of Flow-based Large Diffusion Transformers (Flag-DiT) equipped with zero-initialized attention, as a unified framework designed to transform noise into images, videos, multi-view 3D objects, and audio clips conditioned on text instructions.
- Show, Tell and Discriminate: Image Captioning by Self-retrieval with Partially Labeled Data147
This work proposes an image captioning framework with a self-retrieval module as training guidance, which encourages generating discriminative captions and demonstrates the effectiveness of the proposed retrieval-guided method on COCO and Flickr30k captioning datasets, and shows its superior captioning performance with more discriminating captions.
- End-to-End Deep Kronecker-Product Matching for Person Re-identification147
A novel Kronecker Product Matching module to match feature maps of different persons in an end-to-end trainable deep neural network that outperforms state-of-the-art methods on the Market-1501, CUHK03, and DukeMTMC datasets, which demonstrates the effectiveness and generalization ability of the proposed approach.
- LIF-Seg: LiDAR and Camera Image Fusion for 3D LiDAR Semantic Segmentation146
A coarse-to-fine LiDAR and camera fusion-based network, named LIF-Seg, for LiDAR segmentation, which introduces a simple but effective early-fusion strategy to fully utilize the contextual information of images and can effectively address the weak spatiotemporal synchronization problem.
- Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-Based Perception145
This paper proposes a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern while maintaining these inherent properties.
- Crafting GBD-Net for Object Detection144
This paper proposes a gated bi-directional CNN (GBD-Net) to pass messages among features from different support regions during both feature learning and feature extraction, and shows the effectiveness of different network structures and existing techniques for object detection.
- LIGA-Stereo: Learning LiDAR Geometry Aware Representations for Stereo-based 3D Detector142
This work proposes LIGA-Stereo (LiDAR Geometry Aware Stereo Detector) to learn stereo-based 3D detectors under the guidance of high-level geometry-aware representations of LiDAR-based detection models, and finds existing voxel-based stereo detectors failed to learn semantic features effectively from indirect 3D supervisions.
- Towards Realistic UAV Vision-Language Navigation: Platform, Benchmark, and Methodology135
An assistant-guided UAV object search benchmark called UAV-Need-Help, which provides varying levels of guidance information to help UAVs better accomplish realistic VLN tasks, and a UAV navigation LLM that leverages the multimodal understanding capabilities of the MLLM to jointly process visual and textual information, and performs hierarchical trajectory generation.
- 130
- A Simple Baseline for Video Restoration with Grouped Spatial-Temporal Shift127
The proposed framework is based on grouped spatial-temporal shift, which is a lightweight and straightforward technique that can implicitly capture inter-frame correspondences for multi-frame aggregation and combined with basic 2D convolution, this simple framework can effectively aggregate inter- frame information.
- 127
- 126
- VideoFlow: Exploiting Temporal Cues for Multi-frame Optical Flow Estimation125
VideoFlow concurrently estimates bi-directional optical flows for multiple frames that are available in videos by sufficiently exploiting temporal cues by sufficiently exploiting temporal cues and presents extraordinary performance, ranking 1st on all public benchmarks.
- 121
- 121
- LiDAR-based Panoptic Segmentation via Dynamic Shifting Network121
This work proposes the Dynamic Shifting Network (DS-Net), which serves as an effective panoptic segmentation framework in the point cloud realm and presents an efficient learnable clustering module, dynamic shifting, which adapts kernel functions on-the-fly for different instances.
- 121
- 120
- DigestPath: A benchmark dataset with challenge review for the pathological detection and segmentation of digestive-system117
Two well-annotated benchmark datasets and organized challenges for the digestive-system pathological cell detection and tissue segmentation are released, in conjunction with the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI).
- 117
- MMSearch: Benchmarking the Potential of Large Models as Multi-modal Search Engines113
A delicate pipeline is designed, MMSearch-Engine, to empower any LMMs with multimodal search capabilities, and extensive experiments on closed-source and open-source LMMs are conducted, finding GPT-4o with MMSearch-Engine achieves the best results.
- Encoder-decoder with Multi-level Attention for 3D Human Shape and Pose Estimation112
This work proposes Multi-level Attention Encoder-Decoder Network (MAED), including a Spatial-Temporal Encoder (STE) and a Kinematic Topology Decoder (KTD) to model multi-level attentions in a unified framework.
- Progressive Correspondence Pruning by Consensus Learning112
This work introduces a "pruning" block that lets us identify reliable candidates among the initial matches according to consensus scores estimated using local-to-global dynamic graphs and achieves progressive pruning by stacking multiple pruning blocks sequentially.
- Rethinking Noise Synthesis and Modeling in Raw Denoising109
A new perspective to synthesize noise by directly sampling from the sensor’s real noise is introduced, which inherently generates accurate raw image noise for different camera sensors and demonstrates wide generalization on different sensors and lighting conditions.
- CameraCtrl II: Dynamic Scene Exploration via Camera-Controlled Video Diffusion Models107
This paper builds on improved single-clip capabilities of CameraCtrl II by allowing users to iteratively specify camera trajectories for generating coherent video sequences, and enables extended scene exploration by allowing users to iteratively specify camera trajectories for generating coherent video sequences.
- Lumina-Image 2.0: a Unified and Efficient Image Generative Framework105
An advanced text-to-image (T2I) model that surpasses previous state-of-the-art methods across multiple benchmarks is introduced, and a unified captioning system, Unified Captioner (UniCap), is introduced, which can generate detailed and accurate multilingual captions for the model.
- 104
- Monocular 3D Object Detection with Decoupled Structured Polygon Estimation and Height-Guided Depth Estimation101
A novel unified framework which decomposes the detection problem into a structured polygon prediction task and a depth recovery task is proposed, in which the method achieves state-of-the-art detection accuracy.
- Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understanding100
Lumina-DiMOO achieves state-of-the-art performance on multiple benchmarks, surpassing existing open-source unified multi-modal models and fostering further advancements in multi-modal and discrete diffusion model research.
- Texture-based classification of different single liver lesion based on SPAIR T2W MRI images100
Texture features of SPAIR T2W-MRI can classify the three types of single liver lesions (HH, HM and HCC) and may serve as an adjunct tool for accurate diagnosis of these diseases.
- 99
- Composite splitting algorithms for convex optimization98
The proposed composite splitting algorithms are applied to the compressed MR image reconstruction and low-rank tensor completion and demonstrate the superior performance of the proposed algorithms in terms of both accuracy and computation complexity.
- MixMAE: Mixed and Masked Autoencoder for Efficient Pretraining of Hierarchical Vision Transformers97
Empirical results demonstrate that MixMAE can learn high-quality visual representations efficiently and show that it has better FLOPs / performance tradeoff than previous popular MIM methods.
- Signet Ring Cell Detection with a Semi-supervised Learning Framework97
A semi-supervised learning framework that achieves accurate signet ring cell detection and can be readily applied in the clinical trails and self-training and cooperative-training are proposed to deal with the challenge of incomplete annotations.
- Unsupervised Domain Adaptive 3D Detection with Multi-Level Consistency96
A novel and unified framework, Multi-Level Consistency Network (MLC-Net), which employs a teacher-student paradigm to generate adaptive and reliable pseudo-targets and out-performs existing state-of-the-art methods on standard benchmarks.
- ST3D++: Denoised Self-Training for Unsupervised Domain Adaptation on 3D Object Detection95
ST3D++ achieves state-of-the-art performance on all evaluated settings, outperforming the corresponding baseline by a large margin, and even surpasses the fully supervised oracle results on the KITTI 3D object detection benchmark with target prior.
- DivCo: Diverse Conditional Image Synthesis via Contrastive Generative Adversarial Network95
This paper proposes a novel DivCo framework to properly constrain both "positive" and "negative" relations between the generated images specified in the latent space and introduces a novel latent-augmented contrastive loss, which encourages images generated from adjacent latent codes to be similar and those generated from distinct latent code to be dissimilar.
- Foreground-Action Consistency Network for Weakly Supervised Temporal Action Localization94
A framework named FAC-Net based on the I3D backbone, on which three branches are appended, named class-wise foreground classification branch, class-agnostic attention branch and multiple instance learning branch, to regularize the foreground-action consistency and help to learn a meaningful foreground classifier.
- KBNet: Kernel Basis Network for Image Restoration93
A kernel basis attention (KBA) module is proposed, which introduces learnable kernel bases to model representative image patterns for spatial information aggregation and designs a multi-axis feature fusion block to encode and fuse channel-wise, spatial-invariant, and pixel-adaptive features for image restoration.
- Weakly Supervised Temporal Action Localization via Representative Snippet Knowledge Propagation93
This method seeks to mine the representative snippets in each video for propagating information between video snippets to generate better pseudo labels and obtains superior performance on two benchmarks, THUMOS14 and ActivityNet1.
- FocusNetv2: Imbalanced large and small organ segmentation with adversarial shape constraint for head and neck CT images93
A novel two-stage deep neural network, FocusNetv2, is proposed to solve this challenging problem by automatically locating, ROI-pooling, and segmenting small organs with specifically designed small-organ localization and segmentation sub-networks while maintaining the accuracy of large organ segmentation.
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement92
This paper presents a practical unsupervised image denoising method that only requires single noisy images and a noise model, which is easily accessible in practical raw image Denoising and argues that existing denoised datasets are of low quality and contain only a small number of scenes.
- Auto MC-Reward: Automated Dense Reward Design with Large Language Models for Minecraft89
An advanced learning system that leverages Large Language Models (LLMs) to automatically design dense reward functions, thereby enhancing the learning efficiency of reinforcement-learning-based agents in complex tasks in Minecraft is introduced.
- 88
- Multi-Modality Latent Interaction Network for Visual Question Answering87
The proposed Multi- modality Latent Interaction module (MLI) learns the cross-modality relationships between latent visual and language summarizations, which summarize visual regions and question into a small number of latent representations to avoid modeling uninformative individual region-word relations.
- Deep Continuous Conditional Random Fields With Asymmetric Inter-Object Constraints for Online Multi-Object Tracking87
A deep continuous conditional random field (DCCRF) is proposed for solving the online MOT problem in a track-by-detection framework and is trained in an end-to-end manner for better adapting the influences of visual information as well as inter-object relations.
- RNNPose: Recurrent 6-DoF Object Pose Refinement with Robust Correspondence Field Estimation and Pose Optimization86
This paper proposes a framework based on a recurrent neural network (RNN) for object pose refinement, which is robust to erroneous initial poses and occlusions and introduces a consistency-check mechanism based on the learned descriptors of the 3D model and observed 2D images, which downweights the unreliable correspondences during pose optimization.
- 86
- MM-Interleaved: Interleaved Image-Text Generative Modeling via Multi-modal Feature Synchronizer83
MM-Interleaved is an end-to-end generative model for interleaved image-text data that introduces a multi-scale and multi-image feature synchronizer module, allowing direct access to fine-grained image features in the previous context during the generation process.
- Semantic Scene Completion via Integrating Instances and Scene in-the-Loop83
This work presents a novel framework named Scene-Instance-Scene Network (SISNet), which takes advantages of both in-stance and scene level semantic information, and is capable of inferring fine-grained shape details as well as nearby objects whose semantic categories are easily mixed-up.
- Co-Attending Free-Form Regions and Detections With Multi-Modal Multiplicative Feature Embedding for Visual Question Answering83
This paper proposes a novel deep neural network for VQA that integrates both attention mechanisms and effectively fuses features from free-form image regions, detection boxes, and question representations via a multi-modal multiplicative feature embedding scheme to jointly attend question-related free- form image regions and detection boxes for more accurate question answering.
- Pedestrian Behavior Modeling From Stationary Crowds With Applications to Intelligent Surveillance82
A novel model is proposed to model pedestrian behaviors by incorporating stationary crowd groups as a key component and is demonstrated through multiple applications, including walking path prediction, destination prediction, personality attribute classification, and abnormal event detection.
- Question-Guided Hybrid Convolution for Visual Question Answering81
A novel Question-Guided Hybrid Convolution network for Visual Question Answering (VQA) that applies the group convolution, which consists of question-independent kernels and question-dependent kernels, to reduce the parameter size and alleviate over-fitting.
- RBGNet: Ray-based Grouping for 3D Object Detection79
The RBGNet framework is proposed, a voting-based 3D detector for accurate 3D object detection from point clouds, and a ray-based feature grouping module, which aggregates the point-wise features on object surfaces using a group of determined rays uniformly emitted from cluster centers.
- 78
- Temporal Enhanced Training of Multi-view 3D Object Detector via Historical Object Prediction77
This paper proposes a new paradigm, named Historical Object Prediction (HoP), for multi-view 3D detection to leverage temporal information more effectively and can be easily incorporated into state-of-the-art BEV detection frameworks, including BEVFormer and BEVDet series.
- Crossing-Line Crowd Counting with Two-Phase Deep Neural Networks77
This paper proposes a deep Convolutional Neural Network for counting the number of people across a line-of-interest (LOI) in surveillance videos and shows that the proposed method is robust to variations of crowd density, crowd velocity, and directions of the LOI, and outperforms state- of-the-art LOI counting methods.
- Uni-Perceiver v2: A Generalist Model for Large-Scale Vision and Vision-Language Tasks75
This paper proposes Uni-Perceiver v2, which is the first generalist model capable of handling major large-scale vision and vision-language tasks with competitive performance and proposes an effective optimization technique named Task-Balanced Gradient Normalization to ensure stable multi-task learning with an unmixed sampling strategy, helpful for tasks requiring large batch-size training.
- 74
- From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning73
Experimental results show that ReflectionFlow significantly outperforms naive noise-level scaling methods, offering a scalable and compute-efficient solution toward higher-quality image synthesis on challenging tasks.
- 73
- 73
- Urban Radiance Field Representation with Deformable Neural Mesh Primitives72
The Deformable Neural Mesh Primitive (DNMP) is a flexible and compact neural variant of classic mesh representation, which enjoys both the efficiency of rasterization-based rendering and the powerful neural representation capability for photo-realistic image synthesis.
- 72
- 71
- A Challenger to GPT-4V? Early Explorations of Gemini in Visual Expertise68
A preliminary exploration of Gemini Pro's visual understanding proficiency, which comprehensively covers four domains: fundamental perception, advanced cognition, challenging vision tasks, and various expert capacities, demonstrates the potential of Gemini to be a strong challenger to GPT-4V.
- Engineered Red Blood Cell Biomimetic Nanovesicle with Oxygen Self-Supply for Near-Infrared-II Fluorescence-Guided Synergetic Chemo-Photodynamic Therapy against Hypoxic Tumors67
A red blood cell (RBC) biomimetic theranostic nanovesicle with improved photostability, accumulation of PSs, and oxygen self-supply ability to enhance PDT efficacy upon near-infrared (NIR) laser irradiation is developed.
- 67
- 65
- Multi-organ Segmentation via Co-training Weight-Averaged Models from Few-Organ Datasets65
This paper collaboratively train two networks and let the coupled networks teach each other on un-annotated organs, and co-train weight-averaged models for learning a unified multi-organ segmentation network from few-organ datasets to alleviate the noisy teaching supervisions.
- FocusNet: Imbalanced Large and Small Organ Segmentation with an End-to-End Deep Neural Network for Head and Neck CT Images65
This paper proposes a novel end-to-end deep neural network to solve the problem of imbalanced large and small organ segmentation in head and neck (HaN) CT images by automatically locating, ROI-pooling, and segmenting small organs with specifically designed small-organ sub-networks while maintaining the accuracy of largeorgan segmentation.
- Differentiating brain metastases from different pathological types of lung cancers using texture analysis of T1 postcontrast MR65
The goal of this study was to investigate the feasibility of differentiating brain metastases from different types of lung cancers using texture analysis (TA) of T1 postcontrast MR images.
- 65
- Rectified Diffusion: Straightness Is Not Your Need in Rectified Flow63
It is argued that the success of rectification primarily lies in using a pretrained diffusion model to obtain matched pairs of noise and samples, followed by retraining with these matched noise-sample pairs, and that straightness is not an essential training target for rectification; rather, it is a specific case of flow-matching models.
- CGOF++: Controllable 3D Face Synthesis With Conditional Generative Occupancy Fields63
A new NeRF-based conditional 3D face synthesis framework is proposed, which enables 3D controllability over the generated face images by imposing explicit 3D conditions from 3D face priors and shows more precise 3D controllability than state-of-the-art 2D-based controllable face synthesis methods.
- Structured Domain Adaptation With Online Relation Regularization for Unsupervised Person Re-ID63
This work argues that domain translation has great potential on exploiting valuable source-domain data but the existing methods did not provide proper regularization on the translation process and proposes an end-to-end structured domain adaptation framework with an online relation-consistency regularization term.
- 62
- A2-Net: Molecular Structure Estimation from Cryo-EM Density Volumes62
A deep learning framework for amino acid determination in a 3D Cryo-EM density volume is developed and a sequence-guided Monte Carlo Tree Search (MCTS) to thread over the candidate amino acids to form the molecular structure is designed.
- EfficientFCN: Holistically-Guided Decoding for Semantic Segmentation61
The EfficientFCN is proposed, whose backbone is a common ImageNet pre-trained network without any dilated convolution, and achieves comparable or even better performance than state-of-the-art methods with only 1/3 of the computational cost.
- PATS: Patch Area Transportation with Subdivision for Local Feature Matching60
This paper proposes patch area transportation with Subdivision (PATS), which enables learning scale differences in a self-supervised manner, and can deal with many-to-many relationships in bipartite graph matching.
- VS-Net: Voting with Segmentation for Visual Localization60
A novel prototype-based triplet loss with hard negative mining is proposed, which is able to train semantic segmentation networks with a large number of labels efficiently and can outperform state-of-the-art visual localization methods.
- Generative Adversarial Frontal View to Bird View Synthesis60
The BridgeGAN, a novel generative model for bird view synthesis, is proposed, which is much better than those generated by existing methods, with more consistent global appearance and sharper details.
- Ferroptosis in cancer: revealing the multifaceted functions of mitochondria59
This review delves into the intricate mechanisms governing intracellular ferroptosis, emphasizing the centrality of mitochondria in regulating this process within cancer cells and the potential and hurdles of targeting ferroptosis as a therapeutic avenue to overcome resistance to cancer treatment.
- Mining multi-center heterogeneous medical data with distributed synthetic learning59
This study proposes the Distributed Synthetic Learning (DSL) architecture to learn across multiple medical centers and ensure the protection of sensitive personal information and demonstrates the superior performance of DSL as a high-quality synthetic medical image provider by the use of an ideal synthetic quality metric called Dist-FID.
- 59
- 58
- ManipVQA: Injecting Robotic Affordance and Physically Grounded Information into Multi-Modal Large Language Models57
ManipVQA is proposed, a novel framework that infuses MLLMs with manipulation-centric knowledge through a Visual Question-Answering (VQA) format and preserves the original vision-reasoning abilities while incorporating the newly acquired robotic insights.
- Learnable Histogram: Statistical Context Features for Deep Neural Networks57
A learnable histogram layer is proposed, which learns histogram features within deep neural networks in end-to-end training and is able to back-propagate (BP) errors, learn optimal bin centers and bin widths, and be jointly optimized with other layers in deep networks during training.
- 56
- Trim 3D Gaussian Splatting for Accurate Geometry Representation55
The experimental and theoretical analyses reveal that a relatively small Gaussian scale is a non-negligible factor in representing and optimizing the intricate details in representing and optimizing the intricate details, and the proposed TrimGS maintains relatively small Gaussian scales.
- DetZero: Rethinking Offboard 3D Object Detection with Long-term Sequential Point Clouds51
This work proposes a novel paradigm of offboard 3D object detection, named DetZero, which combines an offline tracker coupled with a multi-frame detector to focus on the completeness of generated object tracks and an attention-mechanism refining module to strengthen contextual information interaction across long-term sequential point clouds for object refining with decomposed regression methods.
- Dynamic Graph Representation Learning for Video Dialog via Multi-Modal Shuffled Transformers51
A semantics-controlled multi-modal shuffled Transformer reasoning framework, consisting of a sequence of Transformer modules, each taking a modality as input and producing representations conditioned on the input question is introduced.
- Lumina-mGPT 2.0: Stand-Alone AutoRegressive Image Modeling50
Lumina-mGPT 2.0 is presented, a stand-alone, decoder-only autoregressive model that revisits and revitalizes the autoregressive paradigm for high-quality image generation and beyond, and is confirmed as a strong, flexible foundation model for unified multimodal generation.
- StereoGAN: Bridging Synthetic-to-Real Domain Gap by Joint Optimization of Domain Translation and Stereo Matching50
The proposed end-to-end training framework introduces two novel losses, i.e., bidirectional multi-scale feature re-projection loss and correlation consistency loss, to help translate all synthetic stereo images into realistic ones as well as maintain epipolar constraints.
- 50
- Generalizable Neural Performer: Learning Robust Radiance Fields for Human Novel View Synthesis49
This work presents a simple yet powerful framework, named Generalizable Neural Performer (GNR), that learns a generalizable and robust neural body representation over various geometry and appearance, and introduces an Implicit Geometric Body Embedding strategy to enhance the robustness based on both parametric 3D human body model and multi-view images hints.
- Categorical Relation-Preserving Contrastive Knowledge Distillation for Medical Image Classification49
A novel Categorical Relation-preserving Contrastive Knowledge Distillation (CRCKD) algorithm, which takes the commonly used mean-teacher model as the supervisor, is proposed, which can distill the relational knowledge more comprehensively.
- Conditional Adversarial Generative Flow for Controllable Image Synthesis49
This paper proposes a novel flow-based generative model named conditional adversarial generative flow (CAGlow), which can synthesize images with conditional information like categories, attributes, and even some unknown properties and ensures the independence of different conditions and outperforms regular Glow to a significant extent.
- 47
- BlinkFlow: A Dataset to Push the Limits of Event-Based Optical Flow Estimation46
A novel simulator, BlinkSim, for the fast generation of large-scale data for event-based optical flow, which improves the generalization performance of state-of-the-art methods by more than 40% on average and up to 90%.
- Learning a Structured Latent Space for Unsupervised Point Cloud Completion46
This work proposes a novel framework, which learns a unified and structured latent space that encoding both partial and complete point clouds, and adopts a series of constraints including structured ranking regularization, latent code swapping constraint, and distribution supervision on the related partial point clouds.
- 46
- Towards Optimal Platform-Based Robot Design for Ankle Rehabilitation: The State of the Art and Future Prospects46
Results show that existing robot-assisted ankle rehabilitation techniques can be classified into wearable exoskeleton and platform-based devices, and that an ideal ankle rehabilitation robot should have aligned rotation center as the ankle joint, appropriate workspace, and actuation torque.
- Fast iteratively reweighted least squares algorithms for analysis-based sparse reconstruction44
A novel algorithm based on the iterative reweighted least squares (IRLS) framework, and is accelerated by the preconditioned conjugate gradient method that achieves superior performance over 14 state‐of‐the‐art algorithms in terms of both accuracy and computational cost.
- GS-DiT: Advancing Video Generation with Pseudo 4D Gaussian Fields through Efficient Dense 3D Point Tracking43
GS-DiT demonstrates strong generalization capabilities and extends the 4D controllability of Gaussian splatting to video generation beyond just camera poses and supports advanced cinematic effects through the manipulation of the Gaussian field and camera intrinsics, making it a powerful tool for creative video production.
- Retrieving-to-Answer: Zero-Shot Video Question Answering with Frozen Large Language Models43
Extensive experiments on several VideoQA benchmarks show that despite with 1.3B parameters and no fine-tuning, the R2A can outperform the 61× larger Flamingo-80B model even additionally trained on nearly 2.1B multi-modal data.
- Large-Scale Generation of Functional and Transplantable Hepatocytes and Cholangiocytes from Human Endoderm Stem Cells43
A large-scale suspension system that uses human endoderm stem cells (hEnSCs) as precursors to generate functional and transplantable hepatocytes or cholangiocytes (E-chos) is described, which reveals that hepatic differentiation of hEn SCs recapitulates in vivo development and that the heterogeneities of the resultant populations can be manipulated by regulating the EGF and MAPK signaling pathways.
- Step-Controlled DPO: Leveraging Stepwise Error for Enhanced Mathematical Reasoning42
This work proposes Step-Controlled DPO (SCDPO), a method for automatically providing stepwise error supervision by creating negative samples of mathematical reasoning rationales that start making errors at a specified step, which can better align the model to understand reasoning errors and output accurate reasoning steps.
- A3VLM: Actionable Articulation-Aware Vision Language Model42
A3VLM, an object-centric, actionable, articulation-aware vision language model, which focuses on the articulation structure and action affordances of objects and can be translated into robot actions using simple action primitives.
- ConQueR: Query Contrast Voxel-DETR for 3D Object Detection42
This paper proposes a simple yet effective sparse 3D detector, named Query Contrast Voxel-DETR (Con-QueR), to eliminate the challenging false positives, and achieve more accurate and sparser predictions, and closes the gap of sparse and dense 3D detectors.
- Decoupled DETR: Spatially Disentangling Localization and Classification for Improved End-to-End Object Detection42
By spatially disentangling the two tasks, this method overcomes the misalignment problem and greatly improves the performance of DETR for object detection and achieves a significant improvement in MSCOCO datasets compared to previous work.
- Open-Edit: Open-Domain Image Manipulation with Open-Vocabulary Instructions42
This work proposes a novel algorithm, named Open-Edit, which is the first attempt on open-domain image manipulation with open-vocabulary instructions, which takes advantage of the unified visual-semantic embedding space pretrained on a general image-caption dataset, and manipulates the embedded visual features by applying text-guided vector arithmetic on the image feature maps.
- P2SGrad: Refined Gradients for Optimizing Deep Face Models42
A novel gradient called P2SGrad (Probability-to-Similarity Gradient), which leverages a cosine similarity instead of classification probability to control the gradients for updating neural network parameters, which makes training process more efficient and faster.
- Landmark matching based retinal image alignment by enforcing sparsity in correspondence matrix42
A novel landmark matching formulation is proposed by enforcing sparsity in the correspondence matrix and offer its solutions based on linear programming and a set of reinforced self-similarities descriptors which can better characterize local photometric and geometric properties of the retinal image.
- MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical Code41
A novel method for generating mathematical code accompanied with corresponding reasoning steps for continued pretraining, leading to a 19.2B-token high-performing mathematical pretraining corpus, which is training several popular base models with this corpus significantly improves their mathematical abilities, leading to the creation of the MathCoder2 family of models.
- 41
- Generalizing Monocular 3D Human Pose Estimation in the Wild40
This paper proposes a principled approach to generate high quality 3D pose ground truth given any in-the-wild image with a person inside, and builds a large-scale dataset, which enables the training of a high quality neural network model, without specialized training scheme and auxiliary loss function, which performs favorably against the state-of theart3D pose estimation methods.
- Pyramid Fusion Transformer for Semantic Segmentation39
This study finds that the per-mask classification decoder on top of a single-scale feature is not effective enough to extract reliable probabilities or masks, and proposes a transformer-based Pyramid Fusion Transformer (PFT) for per-mask approach semantic segmentation with multi-scale features.
- 39
- Magnetic Resonance Fingerprinting with compressed sensing and distance metric learning39
A compressed sensing (CS) framework for simultaneously estimating multiple tissue-related parameters based on the MRF method is proposed and substantially outperforms state-of-the-art methods in terms of accuracy of parameter estimation.
- 39
- A three-stage trajectory generation method for robot-assisted bilateral upper limb training with subject-specific adaptation38
A new three-stage trajectory generation method for bilateral upper limb training using interference analysis that provides guidelines for the standardization of robot-assisted bilateralupper limb training protocols.
- PixWizard: Versatile Image-to-Image Visual Assistant with Open-Language Instructions37
This paper presents a versatile image-to-image visual assistant, PixWizard, designed for image generation, manipulation, and translation based on free-from language instructions that tackles a variety of vision tasks into a unified image-text-to-image generation framework and curates an Omni Pixel-to-Pixel Instruction-Tuning Dataset.
- UE4-NeRF:Neural Radiance Field for Real-Time Rendering of Large-Scale Scene37
This paper proposes a novel neural rendering system called UE4-NeRF, specifically designed for real-time rendering of large-scale scenes at 4K resolution with a frame rate of up to 43 FPS, and demonstrates that the method achieves rendering quality comparable to state-of-the-art approaches.
- Improving Weakly Supervised Temporal Action Localization by Bridging Train-Test Gap in Pseudo Labels36
The pseudo-label generation is formulated as an optimization problem under the constraints in terms of the confidence scores of action instances, and the idea of $\Delta$ pseudo labels is introduced, which enables the model with the ability of self-correction.
- AutoLoss-Zero: Searching Loss Functions from Scratch for Generic Tasks36
This paper proposes AutoLoss-Zero, which is a general framework for searching loss functions from scratch for generic tasks, and designs an elementary search space composed only of primitive mathematical operators to accommodate the heterogeneous tasks and evaluation metrics.
- 36
- FreeSim: Toward Free-viewpoint Camera Simulation in Driving Scenes35
This work proposes FreeSim, a camera simulation method for autonomous driving via 3D Gaussian Splatting and diffusion-based image generation that supports high-quality off-trajectory view synthesis under large deviations of more than 3 meters.
- Segmentation and Vascular Vectorization for Coronary Artery by Geometry-Based Cascaded Neural Network35
A geometry-based cascaded segmentation method for the coronary artery is proposed, which generates an accurate, intact and smooth coronary artery, devoid of any fragmentations of segmented vessels.
- Mimic before Reconstruct: Enhancing Masked Autoencoders with Feature Mimicking35
Mimic before Reconstruct for Masked Autoencoders is proposed, named as MR-MAE, which jointly learns high-level and low-level representations without interference during pre-training, and achieves high-level and low-level accuracy on ImageNet-1K.
- Inverting Generative Adversarial Renderer for Face Reconstruction34
This work introduces a novel Generative Adversarial Renderer (GAR) and proposes to tailor its inverted version to the general fitting pipeline, to tackle the above problem of domain-shift noise in training and optimization.
- EasyRef: Omni-Generalized Group Image Reference for Diffusion Models via Multimodal LLM33
EasyRef is introduced, a novel plug-and-play adaptation method that enables diffusion models to be conditioned on multiple reference images and the text prompt and surpasses both tuning-free methods like IP-Adapter and tuning-based methods like LoRA, achieving superior aesthetic quality and robust zero-shot generalization across diverse domains.
- Semantic Segmentation of Wheat Stripe Rust Images Using Deep Learning33
The results showed that the state-of-art Octave-UNet model can better represent and discern the semantic information over a small region and improve the segmentation accuracy of spores, leaves, and backgrounds in the constructed dataset.
- Learning to Predict Context-Adaptive Convolution for Semantic Segmentation33
A Context-adaptive Convolution Network (CaC-Net) is proposed to predict a spatially-varying feature weighting vector for each spatial location of the semantic feature maps and achieves superior segmentation performance on three public datasets.
- 33
- TrajectoryFormer: 3D Object Tracking Transformer with Predictive Trajectory Hypotheses30
This paper presents TrajectoryFormer, a novel point-cloud-based 3D MOT framework that combines long-term object motion feature and short-term object appearance feature to create per-hypothesis feature embedding, which reduces the computational overhead for spatial-temporal encoding.
- Targeted co-delivery of a photosensitizer and an antisense oligonucleotide based on an activatable hyaluronic acid nanosystem with endogenous oxygen generation for enhanced photodynamic therapy of hypoxic tumors30
By improving the target delivery of PS and ASO, relieving tumor hypoxia, and down-regulation of anti-apoptotic proteins, this HCH@DP nanosystem achieved enhanced PDT efficiency against hypoxic tumors.
- 30
- Unsupervised Cross-Spectral Stereo Matching by Learning to Synthesize30
A novel unsupervised crossspectral stereo matching framework based on image-to-image translation based on cycle consistency and adversarial learning and a novel style adaptation network F-cycleGAN is proposed to improve the robustness of spectral translation.
- 30
- 29
- Statistical Evaluation of No-Reference Image Quality Assessment Metrics for Remote Sensing Images29
This paper reviews 21 commonly employed no-reference image quality metrics (IQMs) and evaluates the robustness of the IQMs based on the criteria of prediction accuracy, prediction monotonicity, and prediction consistency, and selects the IQM with the highest loading coefficient as the representative IQM for that component.
- 29
- Diffusion-NPO: Negative Preference Optimization for Better Preference Aligned Generation of Diffusion Models28
This work argues that existing preference alignment methods neglect the critical role of handling unconditional/negative-conditional outputs, leading to a diminished capacity to avoid generating undesirable outcomes, and proposes a straightforward but versatile effective approach that involves training a model specifically attuned to negative preferences.
- FeatAug-DETR: Enriching One-to-Many Matching for DETRs With Feature Augmentation28
This paper proposes two methods that realize one-to-many matching from a different perspective of augmenting images or image features and significantly accelerates DETR training and boosts detection performance while keeping the inference speed unchanged.
- 28
- Denoising Diffusion Semantic Segmentation with Mask Prior Modeling28
This paper proposes to ameliorate the semantic segmentation quality of existing discriminative approaches with a mask prior modeled by a recently-developed denoising diffusion generative model, and develops a unified architecture that adapts diffusion models for mask prior modeling.
- 27
- 26
- Efficient Burst Raw Denoising with Variance Stabilization and Multi-frequency Denoising Network25
This paper proposes an efficient yet effective burst denoising system with a three-stage design that outperforms state-of-the-art methods, with less computational cost and the low complexity and high-quality performance make deployment on smartphones possible.
- 25
- CodePlot-CoT: Mathematical Visual Reasoning by Thinking with Code-Driven Images24
This work proposes CodePlot-CoT, a code-driven Chain-of-Thought paradigm for thinking with images in mathematics, which leverages the VLM to generate text reasoning as well as executable plotting code, which is then rendered into images as visual thought, to solve mathematical problems.
- PUMA: Empowering Unified MLLM with Multi-Granular Visual Generation24
This work proposes PUMA, emPowering Unified MLLM with Multi-grAnular visual generation, a novel paradigm that tackles the diversity-controllability trade-off and effectively meeting the distinct granularity needs for diverse generation and precise manipulation within a single framework.
- StreamChat: Chatting with Streaming Video24
StreamChat is presented, a novel approach that enhances the interaction capabilities of Large Multimodal Models (LMMs) with streaming video content by innovatively updating the visual context at each decoding step, ensuring that the model utilizes up-to-date video content throughout the decoding process.
- Omnidirectional Information Gathering for Knowledge Transfer-based Audio-Visual Navigation24
ORAN, an omnidirectional audiovisual navigator based on cross-task navigation skill transfer, is presented, an omnidirectional audiovisual navigator based on cross-task navigation skill transfer that significantly outperforms previous competitors.
- Robust Self-Supervised LiDAR Odometry Via Representative Structure Discovery and 3D Inherent Error Modeling24
This letter delves into structure reliability for accurate self-supervised ego-motion estimation and aim to alleviate the influence of unreliable structures in training, inference and mapping phases.
- Ponymation: Learning Articulated 3D Animal Motions from Unlabeled Online Videos23
A new method for learning a generative model of articulated 3D animal motions from raw, unlabeled online videos that decomposes each training video clip into a set of explicit geometric and photometric representations, including a rest-pose 3D shape, an articulated pose sequence, and texture.
- GeoMIM: Towards Better 3D Knowledge Transfer via Masked Image Modeling for Multi-view 3D Understanding23
GeoMIM is a multi-camera vision transformer with Cross-View Attention blocks that uses LiDAR BEV features encoded by the pretrained BEV model as learning targets, achieving state-of-the-art performance for camera-based 3D object detection and 3D segmentation.
- Augmenting atlas-based liver segmentation for radiotherapy treatment planning by incorporating image features proximal to the atlas contours23
A novel atlas-based segmentation technique with inclusion of neighborhood image features through the introduction of a narrow-shell surrounding the target objects was established and suggested that the technique was capable to reliably segment liver cases from CT, 4D-CT and CBCT images with little human interaction.
- Actin Filament Tracking Based on Particle Filters and Stretching Open Active Contour Models23
A novel algorithm for actin filament tracking and elongation measurement is introduced that reduces the PF state spaces to one-dimensional spaces by tracking filament bodies using SOAC and probabilistically estimating tip locations along the curve length of SOACs.
- CoPESD: A Multi-Level Surgical Motion Dataset for Training Large Vision-Language Models to Co-Pilot Endoscopic Submucosal Dissection22
A hierarchical decomposition of ESD motion granularity is designed and a multi-level surgical motion dataset (CoPESD) is introduced for training MLLMs as the robotic Co-Pilot of Endoscopic Submucosal Dissection to comprehend surgical scenarios and reason following surgical robotic motions.
- M3Net: Multimodal Multi-task Learning for 3D Detection, Segmentation, and Occupancy Prediction in Autonomous Driving22
This paper introduces M3Net, a novel multimodal and multi-task network that simultaneously tackles detection, segmentation, and 3D occupancy prediction for autonomous driving and achieves superior performance than single task model.
- 22
- MAVIS: Mathematical Visual Instruction Tuning with an Automatic Data Engine22
This paper proposes MAVIS, the first MA thematical VIS ual instruction tuning paradigm for MLLMs, involving a series of mathematical visual datasets and specialized MLLMs, and introduces MAVIS-Instruct, which is adopted to instruct-tune the MLLM for robust mathematical reasoning skills.
- Urban Architect: Steerable 3D Urban Scene Generation with Layout Prior22
This work introduces Layout-Guided Variational Score Distillation to address model optimization inadequacies, and introduces a compositional 3D layout representation into text-to-3D paradigm, serving as an additional prior.
- ADDP: Learning General Representations for Image Recognition and Generation with Alternating Denoising Diffusion Process22
This method represents the first successful development of general representations applicable to both generation and dense recognition tasks and can be used to generate diverse high-fidelity images and also demonstrate excellent transfer performance on recognition tasks.
- 22
- Hybrid Supervision Learning for Pathology Whole Slide Image Classification22
A hybrid supervision learning framework, when applied in training patch model, can carefully make use of coarse image-level labels to refine generated pixel-level pseudo labels and achieve 5.2% higher specificity than purely training on existing labels while retaining 100% sensitivity, in the task of image- level classification to be positive or negative.
- 22
- MedViLaM: A multimodal large language model with advanced generalizability and explainability for medical data understanding and generation21
This study introduces MedViLaM, a unified vision-language model towards a generalist model for medical data that can flexibly encode and interpret various forms of medical data, including clinical language and imaging, all using the same set of model weights.
- A Unified Multi-Scenario Attacking Network for Visual Object Tracking21
A Unified and Effective Network, named UEN, to attack visual object tracking models, which is able to attack many state-of-the-art trackers effectively and outperform the introduced baseline in terms of attacking ability and attacking efficiency.
- 21
- $L_0$ Regularized Stationary-Time Estimation for Crowd Analysis21
A robust algorithm for estimating how long a foreground pixel becomes stationary is proposed, formulated as an L0 optimization problem, which distinguishes among different foreground objects, which are close or overlapped in the spatio-temporal space by using a locally shared foreground codebook.
- 20
- Simulating Fluids in Real-World Still Images20
This work leverage triangular mesh based on a monocular depth estimator to represent fluid surface layer and simulate the motion with the inspiration of classic physics theory of hybrid Lagrangian-Eulerian method, along with a learnable network so as to adapt to complex real-world image textures.
- 20
- InstructSeq: Unifying Vision Tasks with Instruction-conditioned Multi-modal Sequence Generation19
In this work, InstructSeq is introduced, an instruction-conditioned multi-modal modeling framework that unifies diverse vision tasks through flexible natural language control and handling of both visual and textual data.
- SUG: Single-dataset Unified Generalization for 3D Point Cloud Classification19
Experiments demonstrate that the proposed Single-dataset Unified Generalization (SUG) framework can boost the generalization ability for unseen target domains, even outperforming the existing unsupervised domain adaptation methods that have to access extensive target domain data.
- 19
- Dispersal Range of Anopheles sinensis in Yongcheng City, China by Mark-Release-Recapture Methods19
The results of the present study indicate that local An.
- 19
- 18
- Capivasertib plus fulvestrant in patients with HR-positive/HER2-negative advanced breast cancer: phase 3 CAPItello-291 study extended Chinese cohort18
The benefit-risk profile of capivasertib–fulvestrant in the Chinese cohort was favorable and further exploration in patients with PIK3CA/AKT1/PTEN-non-altered tumors is warranted; further exploration in patients with PIK3CA/AKT1/PTEN-non-altered tumors is warranted.
- RBF-Softmax: Learning Deep Representative Prototypes with Radial Basis Function Softmax18
A novel Radial Basis Function (RBF) distances are introduced to replace the commonly used inner products in the softmax loss function, such that it can adaptively assign losses to regularize the intra-class and inter-class distances by reshaping the relative differences, and thus creating more representative pro-totypes of classes to improve optimization.
- PICABench: How Far Are We from Physically Realistic Image Editing?17
PICABench is introduced, which systematically evaluates physical realism across eight sub-dimension (spanning optics, mechanics, and state transitions) for most of the common editing operations (add, remove, attribute change, etc.), and the PICAEval is proposed, a reliable evaluation protocol that uses VLM-as-a-judge with per-case, region-level human annotations and questions.
- SPP: Sparsity-Preserved Parameter-Efficient Fine-Tuning for Large Language Models17
Results show that SPP significantly enhances the performance of models with different sparsity patterns, especially for those with high sparsity ratios, making it a promising solution for the efficient fine-tuning of sparse LLMs.
- Teach-DETR: Better Training DETR With Teachers17
It is shown that the predicted boxes from teacher detectors are effective medium to transfer knowledge of teacher detectors, which could be either RCNN-based or DETR-based detectors, to train a more accurate and robust DETR model.
- 17
- 17
- Learning Adaptive and Temporally Causal Video Tokenization in a 1D Latent Space16
AdapTok is equipped with a block-wise masking strategy that randomly drops tail tokens of each block during training, and a block causal scorer to predict the reconstruction quality of video frames using different numbers of tokens.
- Learning A Multi-Task Transformer Via Unified And Customized Instruction Tuning For Chest Radiograph Interpretation16
A unified transformer model specifically designed for multi-modal clinical tasks by incorporating customized instruction tuning is demonstrated, which can unify the various vision-intensive tasks in a single training framework with homogeneous model inputs and outputs to increase clinical interpretability in one reading.
- 16
- Resurrect Mask AutoRegressive Modeling for Efficient and Scalable Image Generation15
An improved Bidirectional LLaMA architecture is introduced by replacing causal attention with bidirectional attention and incorporating 2D RoPE, which together form an advanced model, MaskGIL, matching state-of-the-art AR models in the ImageNet 256x256 benchmark, while requiring only 8 inference steps compared to the 256 steps of AR models.
- UniAff: A Unified Representation of Affordances for Tool Usage and Articulation with Vision-Language Models15
This work proposes a comprehensive paradigm, termed UniAff, that integrates 3D object-centric manipulation and task understanding in a unified formulation and constructed a dataset labeled with manipulation-related key attributes, comprising 900 articulated objects and 600 tools from 12 categories.
- UniZero: Generalized and Efficient Planning with Scalable Latent World Models15
This work introduces UniZero, a novel approach that employs a modular transformer-based world model to effectively learn a shared latent space and shows that UniZero significantly outperforms existing baselines in benchmarks that require long-term memory.
- BlinkVision: A Benchmark for Optical Flow, Scene Flow and Point Tracking Estimation Using RGB Frames and Events15
BlinkVision enables extensive benchmarks on three types of correspondence tasks (optical flow, point tracking, and scene flow estimation) for both image-based and event-based methods, offering new observations, practices, and insights for future research.
- 15
- Multiscale registration of medical images based on edge preserving scale space with application in image-guided radiation therapy15
A new multi-scale registration framework called edge preserving multiscale registration (EPMR) was proposed based upon an edge preserving total variation L1 norm (TV-L1) scale space representation, which outperformed other methods in terms of the accuracy and robustness for medical images.
- 14
- Detection of Five Types of HPV Genotypes Causing Anogenital Warts (Condyloma Acuminatum) Using PCR-Tm Analysis Technology14
This work established a high-throughput, simple, low-cost, and accurate HPV-typing assay (polymerase chain reaction-melting temperature [PCR-Tm] analysis) to detect HPV in CA, including single and multiple infections.
- Geometry-Based End-to-End Segmentation of Coronary Artery in Computed Tomography Angiography13
This work proposes a geometry-based end-to-end segmentation method for the coronary artery in computed tomography angiography that generates an accurate, intact and smooth mesh of the coronary artery, devoid of any fragmentations of segmented vessels.
- LiDAR-based 4D Panoptic Segmentation via Dynamic Shifting Network13
This work addresses the task of LiDAR-based panoptic segmentation, which aims to parse both objects and scenes in a unified manner, and proposes the Dynamic Shifting Network (DS-Net), which serves as an effective panopti segmentation framework in the point cloud realm.
- REFINE: Prediction Fusion Network for Panoptic Segmentation13
REFINE, pREdiction FusIon NEtwork for panoptic segmentation is presented, to achieve high-quality panoptIC segmentation by improving cross- task prediction fusion, and within-task prediction fusion.
- A Holistically-Guided Decoder for Deep Representation Learning With Applications to Semantic Segmentation and Object Detection13
This paper proposes one novel holistically-guided decoder which is introduced to obtain the high-resolution semantic-rich feature maps via the multi-scale features from the encoder, and implements the EfficientFCN architecture for semantic segmentation and HGD-FPN for object detection and instance segmentation.
- 13
- 12
- 12
- Silhouette analysis for human action recognition based on maximum spatio-temporal dissimilarity embedding12
A novel manifold embedding method, maximum spatio-temporal dissimilarity embedding, is proposed to embed each action frame into a manifold, where frames from different action classes can be well separated.
- 11
- 11
- Unmasking Bias in Diffusion Model Training11
A simple but effective weighting strategy derived from the unlocked biased part of diffusion models is proposed, which remarkably elevates sample quality and displays improved efficiency in both training and sampling processes, by only adjusting loss weighting strategy.
- 11
- 11
- Stable Consistency Tuning: Understanding and Improving Consistency Models10
A novel framework for understanding consistency models is proposed by modeling the denoising process of the diffusion model as a Markov Decision Process (MDP) and framing consistency model training as the value estimation through Temporal Difference~(TD) Learning.
- ECNet: Effective Controllable Text-to-Image Diffusion Models10
A Spatial Guidance Injector (SGI) is proposed which enhances conditional detail by encoding text inputs with precise annotation information and Diffusion Consistency Loss (DCL), which applies supervision on the denoised latent code at any given time step, enhancing the robustness and accuracy of the output.
- 10
- 10
- 10
- Projection of population structure in China using least squares support vector machine in conjunction with a Leslie matrix model10
The experimental results reveal that the proposed LS†SVM†Leslie model improves the projection accuracy relative to the conventional Leslie matrix model in terms of the percentage error and mean algebraic percentage error.
- Edge and color contexts based object representation and tracking10
An edge-color co-occurrence centroid (ECCC) based object appearance representation is derived by employing the non-uniform quantization of the color space to highlight the object, and canny operator and motion detection to gain the object edges.
- 10
- 9
- 9
- 9
- 9
- Laplace-domain diffuse optical measurement9
This work has proposed a novel method for directly obtaining the Laplace transform data and demonstrated that the use of negative Laplace parameters can provide enhanced sensitivity to perturbations located in deep regions.
- 9
- RelightMaster: Precise Video Relighting with Multi-plane Light Images8
This work proposes RelightMaster, a novel framework for accurate and controllable video relighting, and introduces Multi-plane Light Image (MPLI), a novel visual prompt inspired by Multi-Plane Image (MPI).
- GLID: Pre-training a Generalist Encoder-Decoder Vision Model8
GLID achieves competitive performance on various vision tasks, including object detection, image segmentation, pose estimation, and depth estimation, outper-forming or matching specialist models such as Mask2Former, DETR, ViTPose, and BinsFormer.
- 8
- 8
- 8
- NeuralMarker8
The novel framework NeuralMarker, training a neural network estimating dense marker correspondences under various challenging conditions, is proposed and significantly outperforms previous methods and enables new interesting applications, including Augmented Reality (AR) and video editing.
- Collaboration of Pre-trained Models Makes Better Few-shot Learner8
This paper proposes CoMo, a Collaboration of pre-trained Models that incorporates diverse prior knowledge from various pre-training paradigms for better few-shot learning and introduces a learnable Multi-Knowledge Adapter (MK-Adapter) to adaptively blend the predictions from CLIP and DINO.
- 8
- 8
- Oncogenic Genetic Alterations in Non-Small-Cell Lung Cancer (NSCLC) in Southwestern China8
Yunnan NSCLC patients from Xuanwei and non-Han ethnic patients had an obviously unique prevalence of GAs, and subgroup analysis showed that ALK and ROS1 fusions and rare EGFR mutations were associated with non- Han ethnic patients.
- 8
- 7
- 7
- TerDiT: Ternary Diffusion Models with Transformers7
TerDiT, the first quantization-aware training (QAT) and efficient deployment scheme for extremely low-bit diffusion transformer models, is proposed, demonstrating the feasibility of training extremely low-bit DiT models from scratch while maintaining competitive image generation capacities compared to full-precision models.
- Hippocampus segmentation after brain tumor resection via postoperative region synthesis7
A novel automatic hippocampus segmentation framework via postoperative image synthesis achieves better accuracy for hippocampus segmentation of brain tumor patients who have undergone brain tumor resection and can be used as an automatic contouring tool for hippocampus delineation in hippocampus-sparing radiotherapy.
- Lightweight Network with Variable Asymmetric Rebalancing Strategy for Small and Imbalanced Fault Diagnosis7
A lightweight class imbalanced diagnosis framework based on a depthwise separable Laplace-wavelet convolution network with variable-asymmetric focal loss (DSLWCN-VAFL) and a new asymmetric soft-threshold loss, VAFL, is designed, which reasonably rebalances the contributions of distinct samples during the model training.
- 7
- High dose radiation therapy based on normal tissue constraints with concurrent chemotherapy achieves promising survival of patients with unresectable stage III non-small cell lung cancer7
The significant positive association observed between prescribed dose and survival suggests that individualized isotoxic dose-escalated radiation based on NTC might improve survival in this cohort of stage III NSCLC Chinese patients.
- SAR target recognition based on improved joint sparse representation7
The IJSR model can effectively combine multiple-view SAR images from the same physical target to improve the recognition performance and outperforms state-of-the-art methods, such as the joint sparse representation classification (JSRC) method and the sparse representation classified method (SRC).
- 7
- 7
- Actin Filament Segmentation Using Dynamic Programming7
A novel algorithm for actin filament segmentation in 2D TIRFM image sequences is introduced, where its states are tip locations; given candidate tip locations, actin filaments' body points are inferred by a dynamic programming method, which adaptively generates candidate solutions.
- DirectTriGS: Triplane-based Gaussian Splatting Field Representation for 3D Generation6
DirectTriGS is presented, a novel framework designed for 3D object generation with Gaussian Splatting that effectively encodes both the geometry and texture information, enabling smooth transformation back to Gaussian point clouds and rendering into images by a TriRenderer, with only 2D supervisions.
- TinyLVLM-eHub: Towards Comprehensive and Efficient Evaluation for Large Vision-Language Models6
A comprehensive and efficient evaluation framework, TinyLVLM-eHub, is introduced to assess LVLMs’ performance, including proprietary models, and reveals that closed-source API models like GPT4V and GeminiPro-V excel in most capabilities compared to previous open-source LVLMs, though they show some vulnerability in object hallucination.
- Hybrid3D: learning 3D hybrid features with point clouds and multi-view images for point cloud registration6
This work proposes to learn a 3D hybrid feature that fully exploits the multi-view colored images and point clouds from indoor RGB-D scene scans, and utilizes a novel soft-fusion module to associate and fuse hybrid features in a unified space while alleviating the ambiguities of 2D–3D feature binding.
- Spatial Parsing and Dynamic Temporal Pooling networks for Human-Object Interaction detection6
This paper introduces the Spatial Parsing and Dynamic Temporal Pooling (SPDTP) network, which takes the entire video as a spatio-temporal graph with human and object nodes as input and predicts the difference between interactive and non-interactive pairs through explicit spatial parsing.
- 6
- 6
- 6
- 5
- FlexDrive: Toward Trajectory Flexibility in Driving Scene Gaussian Splatting Reconstruction and Rendering5
An Inverse View Warping technique is introduced to create compact and high-quality images as supervision for the reconstruction of the out-of-path views, enabling high-quality rendering results for those views, where the method outperforms previous methods by a significant margin.
- 5
- SKT: Integrating State-Aware Keypoint Trajectories with Vision-Language Models for Robotic Garment Manipulation5
Experimental results indicate that the VLM-based method significantly enhances keypoint detection accuracy and task success rates, providing a more flexible and general solution for robotic garment manipulation.
- 5
- Bireociclib plus fulvestrant for HR+/HER2- advanced female breast cancer progressed on or after endocrine therapy: phase 3 BRIGHT-2 study interim analysis5
Findings suggest that BF is a promising therapeutic option for patients with HR+/HER2- ABC following ET failure, and while CDK4/6 inhibitors have become standard of care in patients with HR+/HER2- advanced breast cancer (ABC), selecting the optimal inhibitor could improve care further.
- 5
- 5
- 5
- FlowFormer: A Transformer Architecture and Its Masked Cost Volume Autoencoding for Optical Flow5
MCVA enhances FlowFormer by pretraining the cost-volume encoder with a masked autoencoding scheme, which further unleashes the capability of FlowFormer with unlabeled data, and improves FlowFormer all-sided and FlowFormer+MCVA ranks 1st among all published methods on both Sintel and KITTI-2015 benchmarks and achieves the best generalization performance.
- 5
- 5
- 5
- Fault Diagnosis and Maintenance Decision Method of Marine Transformer: A Rough Set Theory Based Study5
A fault diagnosis expert system based on rough set theory, which can better detect ship transformers is described, which will better diagnose and repair the transformer fault.
- Energy-saving Routing Algorithm Based on Cluster in WSN5
This paper takes the improved threshold formula and cluster radios formulas to choose cluster-heads by considering energy-saving, based on typical clustering routing protocol and optimal cluster-head selection formula and proposed the realization of target tracking based on the energy- saving routing algorithm.
- 5
- Progressive simplification and transmission of building polygons based on triangle meshes5
A research framework integrating progressive cartographic generalization and transmission of vector data provides multiple resolution data from coarse to fine as key scales and increments between them which is not available in traditional generalization framework.
- 5
- 4
- IMAGINE-E: Image Generation Intelligence Evaluation of State-of-the-art Text-to-Image Models4
This comprehensive assessment highlights each model's strengths and limitations, particularly the outstanding performance of FLUX.1 and Ideogram2.0 in structured and specific domain tasks, underscoring the expanding applications and potential of T2I models as foundational AI tools.
- 4
- NODI: Out-Of-Distribution Detection with Noise from Diffusion4
This work introduces the diffusion process into the OOD task, and outperforms previous OOD methods across all types of image encoders, and exhibits good robustness with all the image encoders.
- 4
- 4
- Multi-sensor Data Processing and Fusing Based on Kalman Filtering4
Data simulation analysis shows that the use of kalman filtering algorithm for the fusion of the data measured by multi-sensor is to obtain more accurate warehouse target location data, so as to increase the positioning and tracking accuracy of the warehouse target localization and tracking system.
- 4
- 4
- 3
- ProteinAE: Protein Diffusion Autoencoders for Structure Encoding3
ProteinAE is introduced, a novel and streamlined protein diffusion autoencoder designed to overcome challenges by directly mapping protein backbone coordinates from E(3) into a continuous, compact latent space that enables efficient, high-quality structure generation that is competitive with leading structure-based approaches and significantly outperforms prior latent-based methods.
- 3
- 3
- Self-NPO: Data-Free Diffusion Model Enhancement via Truncated Diffusion Fine-Tuning3
This work introduces Self-NPO, a NPO approach that learns exclusively from the model it self, thereby eliminating the need for manual data labeling or reward model training, and demonstrates that it integrates seamlessly into widely used diffusion models, including SD1.5, SDXL, and CogVideoX.
- 3
- GLEAM: Learning to Match and Explain in Cross-View Geo-Localization3
GLEAM-C and GLEAM-X form a comprehensive CVGL pipeline that integrates multi-modal, multi-view alignment with interpretable correspondence analysis, unifying accurate cross-view matching with explainable reasoning and advancing Geo-Localization by enabling models to better Explain And Match.
- UniRL-Zero: Reinforcement Learning on Unified Models with Joint Language Model and Diffusion Model Experts3
This work defines six scenarios for unified model reinforcement learning, providing systematic baselines for reinforcement learning of unified understanding and generation model.
- Revisiting Generative Policies: A Simpler Reinforcement Learning Algorithmic Perspective3
This study compares and analyzes various generative policy training and deployment techniques, identifying and validating effective designs for generative policy algorithms, and introduces a standardized experimental framework named GenerativeRL.
- Enhancing Visual Grounding and Generalization: A Multi-Task Cycle Training Approach for Vision-Language Models3
ViLaM, a unified Vision-Language transformer model that integrates instruction tuning predicated on a large language model, is proposed, enabling it to optimally utilize the knowledge and reasoning capacities of large pre-trained language models for an array of tasks encompassing both language and vision.
- 3
- 3
- 3
- BEAMSPACE BASED DOA ESTIMATION METHODS OF COHERENT SOURCES IN THE PRESENCE OF IMPULSIVE NOISE3
Noise subspace can be estimated by the eigen-decomposition of covariation matrices and FLOM matrices in both element space and beamspace so as to estimate the direction of arrival (DOA) of coherent sources in impulse noise.
- 3
- DSP-based Close-coupled Master-slave Navigation Computer for SINS3
To reduce the size,weight and power consumption of strapdown inertial navigation systems(SINS), a navigation computer featured with closecoupled masterslave dual processors is developed, which employs a digital signal processor as its navigation algorithm processor and a microcontroller as its I/O interface processor.
- 2
- 2
- 2
- 2
- 2
- Bireociclib plus fulvestrant for advanced HR+/HER2- breast cancer progressing after endocrine therapy: Interim analysis of a phase 3 trial (BRIGHT-2).2
Bireociclib plus fulvestrant showed superior efficacy for pretreated HR+/HER2- ABC, especially in patients with primary resistance to ET, liver metastases or bone-only metastases, especially in patients with primary resistance to ET, liver metastases or bone-only metastases.
- 2
- 2
- 2
- Guest Editorial: Generative Adversarial Networks for Computer Vision2
Classic generative models, such as GaussianMixtureModel and principal components, have long been used to learn prior models for image restoration, but none of the above models are able to learn the distribution of in-the-wild natural images and capture long-range dependence beyond local regions.
- MagnifierNet: Towards Semantic Adversary and Fusion for Person Re-identification2
This work proposes MagnifierNet, a triple-branch network which accurately mines details from whole to parts and introduces a novel loss function "Semantic Diversity Loss" to remove redundant overlaps across learned semantic representations.
- 2
- SU‐E‐I‐87: Automated Liver Segmentation Method for CBCT Dataset by Combining Sparse Shape Composition and Probabilistic Atlas Construction2
It is demonstrated that liver structures of CBCT with artifacts can be extracted accurately for following adaptive radiation therapy.
- 2
- [Value of (18)F-FDG PET/CT in the diagnosis of laryngeal carcinoma].2
F-FDG PET/CT can be an important means for diagnosis and initial staging of larynx carcinoma and shows significant differences in the diagnostic sensitivity and accuracy between the two modalities.
- Hybrid deformable image registration using a closed-form Free Form Deformation approach2
This paper describes an efficient non-rigid image registration method using a B-spline based Free Form Deformation (FFD) model that utilises sparse feature correspondences to estimate an elastic deformation field in a closed form.
- 2
- 2
- 1
- 1
- 1
- 1
- Impact of HER2-ultralow heterogeneity and optimal threshold on trastuzumab deruxtecan (T-DXd) efficacy in metastatic breast cancer: A national multicenter cohort study (HEROIC).1
A high HER2-ultralow discordance rate was observed between primary tumors and matched metastases and re-evaluating HER2 status in metastatic lesions is recommended for T-DXd treatment decision.
- 1
- Enhancing Vision-Language Model with Unmasked Token Alignment1
Unmasked Token Alignment (UTA) is introduced, a method that leverages existing CLIP models to further enhance its vision-language representations and outperform existing MIM methods on various uni- and multi-modal benchmarks.
- 1
- https://www.zentimepublishing.com/journal/PMD/issues/Vol1Iss1/10.61189/550253gnnvtv.html1
The technologies listed in this review inculde sedation, acupoint stimulation, recumbent changes, gas-filled assistance, warm water infusion assistance, and distraction work with endoscopy to increase patient tolerance.
- Perception Imitation: Towards Synthesis-free Simulator for Autonomous Vehicles1
In this work, a series of evaluation methods such as matching metric and performance of downstream task are exploited to examine the simulation quality and show that the perception imitation method is effective to model the behavior of learning-based perception model.
- 1
- 1
- Towards Large-scale Masked Face Recognition1
This paper will focus on four challenges in large-scale masked face recognition, i.e., super-large scale training, data noise handling, masked and non-masked face recognition accuracy balancing, and how to design inference-friendly model architecture.
- 1
- Mutational characteristics of Chinese Han and ethnic minorities in southwestern China: a propensity score matched analysis1
This first in-depth large case-control study on genetic mutation profiles among multi-ethnic patients in southwestern China, especially focused on ethnic minorities, may facilitate the understanding of the etiology of this malignant disease and consequently help to reduce the incidence of lung cancer in Yunnan ethnic minority areas.
- 1
- 1
- 1
- 1
- Learning Deep Representations for Scene Labeling with Guided Supervision1
It is argued that the large intra-class variation provides ambiguous training information and hinders the deep models' ability to learn more discriminative deep feature representations.
- 1
- Face Detection Based on Cost-Gentle Adaboost Algorithm1
A novel method is proposed for face detection based on Cost-Gentle Adaboost algorithm that can achieve higher detection rate at appropriate false positive rates and Experimental results show that the method is effective.
- 1
- 1
- 1
- 1
- 1
- 1
- Rapid Algorithm of Building Typification Based on Data Matching1
An algorithm for building typification based on data matching between multi-scale databases is proposed, which supports parallel process and is rapid, effective and adapted to Web mapping.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Application of big data theory in test technology of high precision inertial navigation system–
The problems existing in the big data application of high-precision inertial navigation test technology are explored, the objectives and requirements are analyzed, and the ways and methods of applying big data to high- Precision inertIAL navigation are put forward so as to provide a clear direction for the development ofHigh precision inertial Navigation after shampoo.
- –
- –
- –
- MagnifierNet: Towards Semantic Regularization and Fusion for Person Re-identification.–
A novel network which accurately mines details for each semantic region and selectively fuse all semantic feature representations and introduces a novel loss function “Semantic Diversity Loss” (SD Loss) to facilitate feature diversity and improves regularization among all semantic regions.
- –
- The recent advances in nurses' resilience–
The research progress of nurses' resilience from the definition and classification, evaluation tools, influencing factors and intervention status, and the current situation and development trend are reviewed to point out that further research on nurses’ resilience in China can deepen the research on factors of Nurses' resilience.
- –
- –
- –
- –
- [Research Progress of Quantitative Susceptibility Mapping in MRI].–
The QSM concept is introduced and the various QSM methods are systematically categorized and discussed and the aim of this paper is to summarize the current research progress of QSM, popularize the knowledge of Q SM and promote the improvements and the rational application ofQSM in clinical field.
- –
- –
- Image Scaling Research Based on Associative Memory Algorithm–
The real experimental simulation shows that the proposed algorithm can output images to keep the content of the original image,PSNR large, and the process of image scaling algorithm is shown.
- Infrared image segmentation based on gene evolutionary branch tree algorithm–
Aiming at the characteristics of infrared image segmentation, the gene tree algorithm is proposed and has good performance.
- [Accurate 3D free-form registration between fan-beam CT and cone-beam CT].–
The proposed gradient-based registration algorithm could register more accurately the clinical cone-beam CT and fan- beam CT images compared with the intensity-based algorithm and meets the real-time requirement in the online adaptive radiotherapy.
- Automatic tumour delineation in whole body PET/CT images–
The proposed multi-modal method achieves more accurate and reproducible tumour delineation than using PET or CT alone and is effective for automated delineation of tumour boundaries.
- Mathematical models for object matching and their application to computer vision and biomedical imaging–
This thesis improves the classical Iterative Closest Point method with a robust shape dissimilarity metric and an asymmetric shape representation which allows the objective function to be efficiently evaluated, and proposes a new matching framework supporting all geometric transformation models that can be expressed by convex functions with convex constraints.
- The design of Intelligent Sulfur Measurement System based on P89C668 Single-chip–
The instrument is designed according to the Coulometric titration principle,GB214-2007 " Determination of all-sulfur in coal", and its result is rapid and accurate.
- –
- Deformable Registration Method Using Edge Preserving Scale Space with Application in Adaptive Radiation Therapy–
A multi-scale deformable registration framework is proposed by combining edge preserving scale space with the multi-level free form deformation (FFD) grids and has a promising efficiency when used in CBCT based IGRT system.
- Imaging Diagnosis of Renal Lymphoma–
CT,MRI and PET/CT can clearly demonstrate the lesion of renal lymphoma, which are multiple masses, solitary masses or contiguous retroperitoneal masses extension with homogeneous attenuation and mild enhancement.
- B-spline Numerical Control R Parameter Programming Processing–
The system numerical control milling machine actual processing movement cam transect curve outline section R parameter processing procedure confirmation reduction procedure to be correct and the components precision satisfies the design request.
- –
- –
- Robust Acoustic Source Localization for Digital Hearing Aids in Noise and Reverberant Environment–
A new method in digital hearing aids to adaptively localize the speech source in noise and reverberant environment is proposed, which achieves faster and more accurate convergence in a low signal-to-noise ratio (SNR) environment.
- –
- –
- –
- –
- Grid transformation method of population data combining geographic factors and simulated township boundary adjustment–
The study indicated that the proposed method can acquire fine-grained grid population surface and was demonstrated as an effective method to transform census population data into regular grids.
- Study of defining an appropriate CTV-to-PTV margins for non-small cell lung cancer based on online registration of cone.beam CT images and planning CT images–
There are some extent of errors from positioning and organ motion in 3D-CRT for NSCLC, and online correction approach based on CBCT images analysis can be used to re-duce the impact obviously and to estimate the appropriate margins for the CTV.
- –
- Design of the temperature acquisition module based on Fieldbus–
The design scheme of the temperature acquisition module based on Fieldbus RS485 is introduced, which can gather 6 thermo-resistance signals, and convert them to temperature values based on the MODBUS protocol, and transmit them to industrical computer via the RS485 Fieldbus.
- –
- Digital/Analog ISP and It's EDA Tools–
The features and advantages of ISP are summarized, the current digital and analog programmable devices are discussed and some corresponding EDA tools are introduced.
- –
- Oilfield Remote Monitor System Based on Intranet–
The main technical features and wide prospect of enterprise remote monitor system based on Intranet are introduced and a monitor system scheme oriented to the middle-type and mini-type enterprise is discussed.
- Research on the three-dimensional virtual CNC machine tool software–
The paper discusses S3D CNC machine tool simulation software distinguishing feature that is based on the PC computer and OpenGL graphic interface that can be used for reference in Open Architecture Control System and other CNCmachine tool training software.
- Layered-integrated ANN Diagnosis Models and Applications in Remote Diagnosis Systems–
L layered-Integrated ANN diagnosis models are presented and it proves appropriate to apply the models to remote diagnosis systems through the tests and applications.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.