Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks88 from public data
- A large-scale car dataset for fine-grained categorization and verification882
This paper presents an on-going effort in collecting a large-scale dataset, “CompCars”, that covers not only different car views, but also their different internal and external parts, and rich attributes, and demonstrates a few important applications exploiting the dataset.
- Video Instance Segmentation663
The first time that the image instance segmentation problem is extended to the video domain, and a novel algorithm called MaskTrack R-CNN is proposed for this task, which is simultaneous detection, segmentation and tracking of instances in videos.
- YouTube-VOS: Sequence-to-Sequence Video Object Segmentation598
This work builds a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS) and proposes a novel sequence-to-sequence network to fully exploit long-term spatial-temporal information in videos for segmentation.
- Efficient Video Object Segmentation via Network Modulation373
This work proposes a novel approach that uses a single forward pass to adapt the segmentation model to the appearance of a specific object and is 70× faster than fine-tuning approaches and achieves similar accuracy.
- 334
- 272
- Robust High-Resolution Video Matting with Temporal Guidance226
A robust, real-time, high-resolution human video matting method that achieves new state-of-the-art performance and proposes a novel training strategy that enforces the network on both matting and segmentation objectives, which significantly improves the model’s robustness.
- 216
- Dense Captioning with Joint Inference and Visual Context179
A new model pipeline based on two novel ideas, joint inference and context fusion, is proposed, which achieves state-of-the-art accuracy on Visual Genome for dense captioning with a relative gain of 73% compared to the previous best algorithm.
- $R^{2}$ Former: Unified Retrieval and Reranking Transformer for Place Recognition171
A unified place recognition framework that handles both retrieval and reranking with a novel transformer model, named R^{2}$ Former, which significantly outperforms state-of-the-art methods on major VPR datasets with much less inference time and memory consumption.
- Streamlined Dense Video Captioning171
A novel dense video captioning framework is proposed, which models temporal dependency across events in a video explicitly and leverages visual and linguistic context from prior events for coherent storytelling.
- AdaBits: Neural Network Quantization With Adaptive Bit-Widths155
It is demonstrated that bit-width of weights and activations is a new option for adaptively executable deep neural networks, offering a distinct opportunity for improved accuracy-efficiency trade-off as well as instant adaptation according to the platform constraints in real-world applications.
- FracBits: Mixed Precision Quantization via Fractional Bit-Widths102
A novel learning-based algorithm is proposed to derive mixed precision models end-to-end under target computation constraints and model sizes under different resource constraints on ImageNet dataset.
- Dual Diffusion for Unified Image Generation and Understanding91
A large-scale and fully end-to-end diffusion model for multi-modal understanding and generation that significantly improves on existing diffusion-based multimodal models, and is the first of its kind to support the full suite of vision-language modeling capabilities.
- HR-NAS: Searching Efficient High-Resolution Neural Architectures with Lightweight Transformers82
This work proposes a novel NAS method, called HR-NAS, which is able to find efficient and accurate networks for different tasks, by effectively encoding multiscale contextual information while maintaining high-resolution representations.
- CompFeat: Comprehensive Feature Aggregation for Video Instance Segmentation80
This work proposes a novel comprehensive feature aggregation approach (CompFeat) to refine features at both frame-level and object-level with temporal and spatial context information to eliminate ambiguities introduced by only using single-frame features.
- Progressive Temporal Feature Alignment Network for Video Inpainting72
This work proposes a ‘Progressive Temporal Feature Alignment Network’, which progressively enriches features extracted from the current frame with the feature warped from neighbouring frames using optical flow, greatly improving visual quality and temporal consistency of the inpainted videos.
- Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model67
This work presents Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering.
- Neural Architecture Search for Lightweight Non-Local Networks55
This work proposes a Lightweight Non-Local (LightNL) block by squeezing the transformation operations and incorporating compact features, and proposes an efficient neural architecture search algorithm to learn an optimal configuration of LightNL blocks in an end-to-end manner.
- 37
- Why Is Prompt Tuning for Vision-Language Models Robust to Noisy Labels?35
It is demonstrated that noisy zero-shot predictions from CLIP can be used to tune its own prompt, significantly enhancing prediction accuracy in the unsupervised setting.
- 34
- 33
- Context-Aware Zero-Shot Recognition32
This work aims to understand the identity of the novel objects in an image surrounded by the known objects using the inter-object relation prior, and leverage the visual context and the geometric relationships between all pairs of objects in a single image, and capture the information useful to infer unseen categories.
- Finetuned Multimodal Language Models Are High-Quality Image-Text Data Filters31
This work proposes a novel framework for filtering image-text data by leveraging fine-tuned Multimodal Language Models (MLMs), which outperforms predominant filtering methods and achieves significant improvements over CLIPScore on popular foundation models and various downstream tasks.
- 28
- Object Detection and Viewpoint Estimation with Auto-masking Neural Network28
This study has designed a novel deep architecture, called Auto-masking Neural Network (ANN), for object detection and viewpoint estimation that can automatically learn to select the most discriminative object parts across different viewpoints from training images.
- Automatic relative radiometric normalization using iteratively weighted least square regression28
This paper presents a new automatic normalization approach that uses the iteratively weighted least square regression technique, and is robust to outliers since it adaptively places different weights on different pixels according to their probabilities of no‐change.
- VMAs: Video-to-Music Generation via Semantic Alignment in Web Music Videos25
This work develops a generative video-music Transformer with a novel semantic video-music alignment scheme that outperforms existing approaches on the DISCO-MV and MusicCaps datasets according to various music generation evaluation metrics, including human evaluation.
- The Devil is in the Details: A Deep Dive into the Rabbit Hole of Data Filtering24
This paper describes the learning and solution when participating in the DataComp challenge, and integrates existing methods and proposes new solutions, such as computing CLIP score on horizontally flipped images to mitigate the interference of scene text.
- 23
- Autonomous UAVs landing site selection from point cloud in unknown environments21
This paper presents a novel and complete framework for UAVs landing, which is built on point cloud in coarse-to-fine manner and has four modules: point cloud preprocessing, coarse landing site selection, fine terrain evaluation, and landing optimal model.
- 19
- Dynamic Proposals for Efficient Object Detection18
This paper designs a module to make a single query-based model to be able to inference with different numbers of proposals, and extends it to a dynamic model to choose the number of proposals according to the input image, greatly reducing computational costs.
- 17
- Video-Teller: Enhancing Cross-Modal Generation with Fusion and Decoupling16
Experimental results demonstrate the efficacy of the proposed video-language foundation model in accurately comprehending videos and generating coherent and precise language descriptions and the fine-grained alignment enhances the model's capabilities with only 13% extra parameters in training and zero additional cost in inference.
- Neural Network Quantization with Scale-Adjusted Training15
This paper proposes a technique called scale-adjusted training (SAT) by directly scaling down weights in the last layer to alleviate over-fitting problem and achieves state-of-the-art accuracy with consistent improvement over previous quantization methods.
- 14
- 13
- Revisiting Training-free NAS Metrics: An Efficient Training-based Method13
This paper proposes a light-weight training-based metric which has a weak correlation with the #Param while achieving better performance than training-free metrics at a lower search cost and is competitive among state-of-the-art NAS methods.
- Learning Dynamic Query Combinations for Transformer-based Object Detection and Segmentation13
This work empirically finds that random convex combinations of the learned queries are still good for the corresponding models, and proposes to learn a convex combination with dynamic coefficients based on the high-level semantics of the image to better capture the prior of object locations and categories in the different images.
- 13
- 12
- 12
- AutoSpace: Neural Architecture Search with Less Human Interference11
A novel differentiable evolutionary framework named AutoSpace is proposed, which evolves the search space to an optimal one with following novel techniques: a differentiable fitness scoring function to efficiently evaluate the performance of cells and a reference architecture to speedup the evolution procedure and avoid falling into sub-optimal solutions.
- Bag of shape descriptor using unsupervised deep learning for non-rigid shape recognition11
A bag of shape descriptor based on unsupervised deep learning and Bag of Words (BoW) for shape recognition that effectively overcomes obstacles, such as irregular topology, orientation ambiguity, and rigid or non-rigid transformation in the hierarchical learning of contour fragments.
- Depth From Water Reflection11
This paper designs a framework to reconstruct the depth from a single water reflection image, and proposes an appearance adaptation method to transform the appearance of the mirror scene so that it is much closer to the real scene.
- Video Recognition in Portrait Mode10
This paper has developed the first dataset dedicated to portrait mode video recognition, namely PortraitMode-400, comprising 400 fine-grained categories, and rigorous quality assurance was implemented to ensure the accuracy of human annotations.
- Dual-Sampling Attention Pooling for Graph Neural Networks on 3D Mesh10
Uniform sampling was first used in this study to construct a multi-scale mesh hierarchy and vertex clustering sampling was used because it can preserve the geometric structure, especially the edge information, and dual-sampling attention pooling for graph neural networks on 3D mesh is proposed.
- Bag of feature with discriminative module for non-rigid shape retrieval10
A novel shape retrieval method, namely BoF-DM (Bag of feature with Discriminative Module), which is specially designed to highlight discriminative features from the multi-scale context structures and fully considers the attention issue of inter-class difference and intra-class similarity.
- Exploring the Role of Audio in Video Captioning8
This work presents an audio-visual framework, which aims to fully exploit the potential of the audio modality for captioning, and proposes new local-global fusion mechanisms to improve information exchange across audio and video.
- 7
- Is In-Domain Data Really Needed? A Pilot Study on Cross-Domain Calibration for Network Quantization7
This work goes beyond the domain of natural images to include drastically different domains such as X-ray images, satellite images and ultrasound images, and finds cross-domain calibration leads to surprisingly stable performance of quantized models on 10 tasks in different image domains with 13 different calibration datasets.
- 7
- Fast Prompt Alignment for Text-to-Image Generation6
This paper introduces Fast Prompt Alignment (FPA), a prompt optimization framework that leverages a one-pass approach, enhancing text-to-image alignment efficiency without the iterative overhead typical of current methods like OPT2I.
- A multi-workpieces recognition algorithm based on shape-SVM learning model6
To make the training recognition algorithm free from the influence of workpieces dimension and rotation, SVH is created to express and wrap the contour features, and SVM is adopted to complete the training and prediction of workpiece type in this paper.
- End-to-End Autoregressive Image Generation with 1D Semantic Tokenizer5
This work designs an end-to-end training pipeline that jointly optimizes reconstruction and generation, enabling direct supervision from generation results to the tokenizer, contrasts with prior two-stage approaches that train tokenizers and generative models separately.
- Weakly Supervised Body Part Parsing with Pose based Part Priors.5
This work proposes the first weakly supervised body part parsing framework and shows that the proposed framework can be extended to other keypoint-supervised part parsing tasks such as face parsing.
- 4
- 4
- 4
- An Automatic Fault Recognition Method for Side Frame Key in TFDS4
An automatic fault recognition method for side frame key in TFDS is proposed based on open source computer vision library (OpenCV) to overcome the disadvantages of manual fault recognition.
- Local projection-based character segmentation method for historical Chinese documents4
The proposed character segmentation algorithm is evaluated based on matching degrees of character bounding boxes between segmentation results and ground-truth data, and achieves a recall rate of 74.3% on a test set, which shows the effectiveness of the proposed algorithm.
- 3
- An Intelligent Defect Detection Method of Small Sized Ceramic Tile Using Machine Vision3
An intelligent detection method is proposed that has good detection effect and high real-time performance and the optimized parameters from feature distribution and a discriminant function are utilized to achieve target defect detection.
- 2
- 2
- 2
- 2
- Grant-Free Transmission by LDPC Matrix Mapping and Integrated Cover-MPA Detector2
A novel transceiver architecture is proposed to simultaneously achieve efficient random access and reliable data transmission in massive IoT networks to realize a practical one-step grant-free transmission and has a compelling performance.
- Rethinking Neural Network Quantization2
Through SAT, quantized models obtain comparable or even better performance than their full-precision counterparts, achieving state-of-the-art accuracy with consistent improvement over previous quantization methods on a wide spectrum of models.
- 2
- 2
- 1
- 1
- Selective Feature Adapter for Dense Vision Transformers1
The experiments show that the dual adapter module, a.k.a SFA, is essential to achieve the best trade-off on dense vision tasks, such as segmentation, detection and depth-estimation, outperforming other adapters with a single module.
- Fault sample augmentation method: Integrating adaptive signal reconstruction and multi-scale stochastic quantization–
Experimental results on multiple benchmark datasets demonstrate that the proposed method significantly improves diagnostic performance and achieves robust generalization under imbalanced conditions.
- –
- Research on Equipment Fault Diagnosis Method Based on Multimodal Deep Fusion–
Compared with existing approaches, the proposed method exhibits superior performance in multi-class recognition, generalization, and anti-interference capability, providing an efficient solution for intelligent fault diagnosis of industrial equipment.
- –
- –
- –
- –
- –
- Improved Sparse Vector Code Based on Optimized Spreading Matrix for Short-Packet URLLC in mMTC–
Simulation results reveal that, compared with the existing work, the block error rate (BLER) performance of SVC can be improved significantly with the optimized spreading matrices.
- –
- –
- Image Processing in DPIV Based on Mathematical Morphology–
It is shown that image noise removing method based on Mathematical Morphology is an effective preprocess method.
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 2026-10-15. 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.