Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Ashraf Ali KassimSuggested from co-authorship
- Shuicheng YanSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works267 from public data
- MedMNIST v2 - A large-scale lightweight benchmark for 2D and 3D biomedical image classification1,572
Covering primary data modalities in biomedical images, MedMNIST v2 is designed to perform classification on lightweight 2D and 3D images with various dataset scales (from 100 to 100,000) and diverse tasks (binary/multi-class, ordinal regression, and multi-label).
- HCP: A Flexible CNN Framework for Multi-Label Image Classification784
Experimental results on Pascal VOC 2007 and VOC 2012 multi-label image datasets well demonstrate the superiority of the proposed HCP infrastructure over other state-of-the-arts, where an arbitrary number of object segment hypotheses are taken as the inputs.
- Adversarial Domain Adaptation with Domain Mixup520
This paper presents adversarial domain adaptation with domain mixup (DM-ADA), which guarantees domain-invariance in a more continuous latent space and guides the domain discriminator in judging samples' difference relative to source and target domains.
- Crowded Scene Analysis: A Survey488
The background knowledge and the available features related to crowded scenes are provided and existing models, popular algorithms, evaluation protocols, and system performance are provided corresponding to different aspects of the crowded scene analysis.
- MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis475
MedMNIST Classification Decathlon is designed to benchmark AutoML algorithms on all 10 datasets, and has compared several baseline methods, including open-source or commercial AutoML tools.
- Modeling Point Clouds With Self-Attention and Gumbel Subset Sampling455
This work develops Point Attention Transformers (PATs), using a parameter-efficient Group Shuffle Attention (GSA) to replace the costly Multi-Head Attention, and proposes an end-to-end learnable and task-agnostic sampling operation, named Gumbel Subset Sampling (GSS), to select a representative subset of input points.
- 397
- 373
- 364
- 361
- Unsupervised Deep Learning for Optical Flow Estimation327
This work devise a simple yet effective unsupervised method for learning optical flow, by directly minimizing photometric consistency by using image warping by the estimated flow.
- 292
- Cross-Domain Detection via Graph-Induced Prototype Alignment284
A Graph-induced Prototype Alignment framework to seek for category-level domain alignment via elaborate prototype representations through graph-based information propagation among region proposals, and in order to alleviate the negative effect of class-imbalance on domain adaptation, a Class-reweighted Contrastive Loss is designed to harmonize the adaptation training process.
- Person Re-identification via Recurrent Feature Aggregation253
This work shows that a progressive/sequential fusion framework based on long short term memory (LSTM) network aggregates the frame-wise human region representation at each time stamp and yields a sequence level human feature representation.
- 3D Human Action Representation Learning via Cross-View Consistency Pursuit248
CrosSCLR consists of both single-view contrastive learning (Skeleton-CLR) and cross-view consistent knowledge mining (CVC-KM) modules, integrated in a collaborative learning manner, and achieves remarkable action recognition results on NTU-60 andNTU-120 datasets under unsupervised settings.
- 246
- Deep Regression Tracking with Shrinkage Loss243
This work proposes a novel shrinkage loss to penalize the importance of easy training data and applies residual connections to fuse multiple convolutional layers as well as their output response maps to balance training data.
- 3D Deep Learning from CT Scans Predicts Tumor Invasiveness of Subcentimeter Pulmonary Adenocarcinomas223
A deep learning system based on 3D convolutional neural networks and multitask learning, which automatically predicts tumor invasiveness, together with 3D nodule segmentation masks is developed, which could help doctors work efficiently and facilitate the application of precision medicine.
- Omni Aggregation Networks for Lightweight Image Super-Resolution211
Two enhanced components under a new Omni-SR architecture are proposed based on dense interaction principle, which can simultaneously model pixel-interaction from both spatial and channel dimensions, and a multi-scale interaction scheme is proposed to mitigate sub-optimal ERF in shallow models.
- Deep-learning-assisted detection and segmentation of rib fractures from CT scans: Development and validation of FracNet211
The proposed FracNet provided increasing detection sensitivity of rib fractures with significantly decreased clinical time consumed, which established a clinically applicable method to assist the radiologist in clinical practice.
- Learning Context Graph for Person Search209
This work proposes a contextual instance expansion module, which employs a relative attention module to search and filter useful context information in the scene, and builds a graph learning framework to effectively employ context pairs to update target similarity.
- 205
- 194
- 170
- 169
- 169
- 160
- Video Summarization via Semantic Attended Networks154
A semantic attended video summarization network (SASUM) which consists of a frame selector and video descriptor to select an appropriate number of video shots by minimizing the distance between the generated description sentence of the summarized video and the human annotated text of the original video.
- Learning to photograph150
An intelligent photography system, which automatically and professionally generates/recommends user-favorite photo(s) from a wide view or a continuous view sequence, and can function as intelligent professionalview guider based on real-time view quality assessment and the embedded compass.
- Learning to Combine: Knowledge Aggregation for Multi-source Domain Adaptation137
A graph model is constructed on the prototypes of various domains to realize the information propagation among semantically adjacent representations and a Relation Alignment Loss (RAL) is designed to facilitate the consistency of categories' relational interdependency and the compactness of features, which boosts features' intra-class invariance and inter-class separability.
- Toward automatic prediction of EGFR mutation status in pulmonary adenocarcinoma with 3D deep learning126
The proposed deep learning system predicts EGFR‐mutant of lung adenocarcinomas in CT images noninvasively and automatically, indicating its potential to help clinical decision‐making by identifying eligible patients of pulmonary adenOCarcinoma for EG FR‐targeted therapy.
- Web image mining towards universal age estimator125
An automatic web image mining system towards building a universal human age estimator based on facial information, which is applicable to all ethnic groups and various image qualities and is robust to the noises existing in both Internet images and corresponding age labels.
- CiaoSR: Continuous Implicit Attention-in-Attention Network for Arbitrary-Scale Image Super-Resolution123
This paper explicitly design an implicit attention network to learn the ensemble weights for the nearby local features in a continuous implicit attention-in-attention network, called CiaoSR, which significantly outperforms the existing single image SR methods with the same backbone.
- 118
- 117
- Performance Guaranteed Network Acceleration via High-Order Residual Quantization112
This paper proposes a highorder binarization scheme, which achieves more accurate approximation while still possesses the advantage of binary operation, and recursively performs residual quantization and yields a series of binary input images with decreasing magnitude scales.
- 111
- 110
- HCSC: Hierarchical Contrastive Selective Coding98
This work proposes a novel contrastive learning framework called Hierarchical Contrastive Selective Coding (HCSC), in which a set of hierarchical prototypes are constructed and also dynamically updated to represent the hierarchical semantic structures underlying the data in the latent space.
- Skeleton-Aided Articulated Motion Generation96
This work makes the first attempt to generate articulated human motion sequence from a single image by utilizing paired inputs including human skeleton information as motion embedding and a single human image as appearance reference to generate novel motion frames based on the conditional GAN infrastructure.
- 93
- A multi-omics-based serial deep learning approach to predict clinical outcomes of single-agent anti-PD-1/PD-L1 immunotherapy in advanced stage non-small-cell lung cancer.87
The SimTA-based multi-omics serial deep learning provides a promising methodology for predicting response of advanced NSCLC patients to anti-PD-1/PD-L1 monotherapy and could better differentiate survival benefit among SD patients than the traditional RECIST evaluation method.
- 76
- 74
- 74
- Learning Multi-Attention Context Graph for Group-Based Re-Identification71
A novel unified framework based on graph neural networks is proposed to simultaneously address the above two group-based re-id tasks, i.e., group re-ID and group-aware person re-ids, and a context graph with group members as its nodes to exploit dependencies among different people is constructed.
- Geometric Constrained Joint Lane Segmentation and Lane Boundary Detection71
A multiple-task learning framework to segment lane areas and detect lane boundaries simultaneously and achieves the best performance on the metrics and a robust detection in varied traffic scenes.
- 69
- Cross-category Video Highlight Detection via Set-based Learning68
A Dual-Learner-based Video Highlight Detection (DL-VHD) framework to improve the conventional pair-based learning by assessing the highlight extent of a video segment under a broader context and outperforms five typical Unsupervised Domain Adaptation (UDA) algorithms on various cross-category highlight detection tasks.
- 66
- 64
- Sense beauty via face, dressing, and/or voice63
This work performs a comprehensive study of female attractiveness conveyed by single/multiple modalities of cues, i.e., face, dressing and/or voice, and aims to uncover how different modalities individually and collectively affect the human sense of beauty.
- 61
- 61
- 61
- 59
- 59
- 58
- Bilevel Online Adaptation for Out-of-Domain Human Mesh Reconstruction57
A new training algorithm named Bilevel Online Adaptation (BOA) is proposed, which divides the optimization process of overall multi-objective into two steps of weight probe and weight update in a training iteration, and leads to state-of-the-art results on two human mesh reconstruction benchmarks1.
- 57
- 54
- 49
- Context-Aware Image Inpainting with Learned Semantic Priors48
A context-aware image inpainting model, which adaptively integrates global semantics and local features in a unified image generator that achieves the state of the art on Places2, CelebA, and Paris StreetView datasets is proposed.
- Exploring Visual Context for Weakly Supervised Person Search45
This paper inventively considers weakly supervised person search with only bounding box annotations, and discovers three levels of context clues in unconstrained natural images that promote local and global discriminative capabilities and enhances clustering accuracy.
- 45
- Generalized Deep 3D Shape Prior via Part-Discretized Diffusion Process44
A vector quantized variational autoencoder is utilized to index local geometry from a compactly learned code-book based on a broad set of task training data, and a discrete diffusion generator is introduced to model the inherent structural dependencies among different tokens.
- Order-Preserving Sparse Coding for Sequence Classification44
This work introduces a temporal-order-preserving regularizer which aims to preserve the temporal order of the reconstruction coefficients, and an efficient Nesterov-type smooth approximation method is developed for optimization of the new regularization criterion, with guaranteed error bounds.
- 43
- 43
- 42
- Out-of-Domain Human Mesh Reconstruction via Dynamic Bilevel Online Adaptation42
The Dynamic Bilevel Online Adaptation algorithm (DynaBOA) first introduces the temporal constraints to compensate for the unavailable 3D annotations and leverages a bilevel optimization procedure to address the conflicts between multi-objectives.
- Egocentric Activity Prediction via Event Modulated Attention42
This work explicitly addresses issues in state-of-the-art egocentric activity understanding techniques by proposing an asynchronous gaze- event driven attentive activity prediction network, built on a gaze-event extraction module inspired by the fact that gaze moving in/out of a certain object most probably indicates the occurrence/ending of acertain activity.
- 41
- 41
- ImplicitAtlas: Learning Deformable Shape Templates in Medical Imaging40
The proposed framework, named ImplicitAtlas, represents a shape as a deformation field from a learned template field, where multiple templates could be integrated to improve the shape representation at negligible computational cost.
- RibSeg Dataset and Strong Point Cloud Baselines for Rib Segmentation from CT Scans40
A labeled rib segmentation benchmark, named RibSeg, is developed, including 490 CT scans (11,719 individual ribs) from a public dataset, and a point cloud-based baseline method is designed, achieving state-of-the-art segmentation performance with significant efficiency.
- 39
- 38
- Action Recognition With Coarse-to-Fine Deep Feature Integration and Asynchronous Fusion38
This paper introduces a coarse-to-fine network which extracts shared deep features at different action class granularities and progressively integrates them to obtain a more accurate feature representation for input actions, and introduces an asynchronous fusion network which fuses information from different streams by asynchronously integrating stream-wise features in different time points.
- Recognition Oriented Facial Image Quality Assessment via Deep Convolutional Neural Network38
A novel facial image quality automatic assessment framework directly targeting on "selecting better face image for better face recognition", which is able to sort the input face images accordingly and "select" good face images for recognition.
- Hierarchical Style-Based Networks for Motion Synthesis37
This paper proposes a self-supervised method for generating long-range, diverse and plausible behaviors to achieve a specific goal location through bi-linear transformation modelling, and demonstrates the generated sequences are useful as subgoals for actual physical execution in the animated world.
- Prediction of future imagery of lung nodule as growth modeling with follow-up computed tomography scans using deep learning: a retrospective cohort study34
A CT-based visual forecasting system that can visualize and quantify a nodule in three dimensions (3D) in any future time point using follow-up CT scans is developed and demonstrated that the deep learning-based model can accurately forecast the imagery of a nodules in a given future for both GGNs and solid nodules.
- Relational Learning Between Multiple Pulmonary Nodules via Deep Set Attention Transformers33
This study proposes a multiple instance learning (MIL) approach and empirically proves the benefit to learn the relations between multiple nodules, and introduces a self-attention transformer equipped with 3D CNN, named NoduleSAT, to replace typical pooling-based aggregation in multiple instancelearning.
- 33
- Learning transform-aware attentive network for object tracking33
A transform-aware attentive tracking framework, which uses a deep attentive network to directly predict the target states via spatial transform parameters, built on the Spatial Transform Network (STN).
- 32
- Asymmetric 3D Context Fusion for Universal Lesion Detection31
A novel asymmetric 3D context fusion operator (A3D), which uses different weights to fuse3D context from different 2D slices that consistently outperforms symmetric context fusion operators by considerable margins, and establishes a new state of the art on DeepLesion.
- Live Face Verification with Multiple Instantialized Local Homographic Parameterization31
This method directly explores the underlying subtle motion difference between the facial movements re-captured from a planer screen and those from a real face; therefore interactive facial expression is no longer required.
- Graphical Modeling for Multi-Source Domain Adaptation30
Two types of graphical models are proposed, i.e. Conditional Random Field for MSDA (CRF-MSDA) and Markov Random Field for MSDA (MRF-MSDA), for cross-domain joint modeling and learnable domain combination, which are more expressive and possesses lower computational cost.
- 30
- 29
- Video Prediction via Selective Sampling29
This work proposes a two-stage network called VPSS, which produces a collection of high quality proposals, facilitated by a multiple choice adversarial learning scheme, yielding diverse frame proposal set.
- Building descriptive and discriminative visual codebook for large-scale image applications29
This paper gives a comprehensive survey on visual vocabulary and details several state-of-the-art algorithms on descriptive and discriminative visual word generation, and introduces two interesting post processing strategies to further improve the performance of visual vocabulary.
- RibSeg v2: A Large-Scale Benchmark for Rib Labeling and Anatomical Centerline Extraction28
This work extends their prior dataset (RibSeg) on the binary rib segmentation task to a comprehensive benchmark, named RibSeg v2, and develops a pipeline including deep learning-based methods for rib labeling, and a skeletonization-based method for centerline extraction.
- 28
- Multi-level attention model for person re-identification28
A feature attention block is proposed for person re-identification that can be seamlessly integrated into existing CNN structures, and is trained only with identity loss.
- Multitask Low-Rank Affinity Graph for Image Segmentation and Image Annotation28
This article investigates a low-rank representation--based graph, which can used in graph-based vision tasks including image segmentation and image annotation, and aims at inferring a unified affinity matrix that implicitly encodes the relations among these patches.
- 27
- Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds27
A novel learning scheme named Self-Prediction is developed for 3D instance and semantic segmentation of point clouds to enhance point relation exploring for better segmentation and achieves state-of-the-art instance segmentation results on S3DIS and comparable semantic segmentsation results when the authors only take PointNet++ as the backbone network.
- Learning explicit video attributes from mid-level representation for video captioning27
Experimental results on video captioning tasks show that the proposed method, utilizing only RGB frames as input without extra video or text training data, could achieve competitive performance with state-of-the-art methods.
- Fully context-aware image inpainting with a learned semantic pyramid26
The Semantic Pyramid Network (SPN) is motivated by the idea that learning multi-scale semantic priors from specific pretext tasks can greatly benefit the recovery of locally missing content in images.
- Sketch Generation with Drawing Process Guided by Vector Flow and Grayscale26
A novel image-to-pencil translation method that could not only generate high-quality pencil sketches but also offer the drawing process, and is superior to others in terms of texture quality, style, and user evaluation.
- Efficient Quantization for Neural Networks with Binary Weights and Low Bitwidth Activations26
This paper designs a new activation function dubbed CReLU from the quantization perspective and further complement this design with appropriate initialization method and training procedure and develops a specific quantization strategy in which the forward and backward approximation of weights with binary values and quantize the activations to low bitwdth using linear or logarithmic quantizer.
- 25
- 25
- 25
- On random hyper-class random forest for visual classification25
A random hyper-class random forest for visual classification tasks (RHC-RF) is developed that well preserves the randomness by random partition of the label space, and at the same time, the objective of training each node directly targets at discriminating classes in a divide-and-conquer-like manner.
- 25
- 24
- Robust geometric ℓ p -norm feature pooling for image classification and action recognition24
This paper proposes to generalize previous pooling methods toward a weighted źp-norm spatial pooling function tailored for class-specific feature spatial distribution, and proposes a simple yet effective self-alignment step during both learning and testing to adaptively adjust the pooling weights for individual images.
- 24
- Deep Rib Fracture Instance Segmentation and Classification From CT on the RibFrac Challenge23
The analysis revealed that several top rib fracture detection solutions achieved performance comparable or even better than human experts, and lay a foundation for future research and development in AI-assisted rib fracture diagnosis.
- GMILT: A Novel Transformer Network That Can Noninvasively Predict EGFR Mutation Status23
A novel explainable transformer network for EGFR classification named gated multiple instance learning transformer (GMILT) is created by integrating multi-instance learning and discriminative weakly supervised feature learning by integrating multi-instance learning and discriminative weakly supervised feature learning.
- Omni-GAN: On the Secrets of cGANs and Beyond23
Omni-GAN is presented, a variant of cGAN that reveals the devil in designing a proper discriminator for training the model, and can extrapolate low-resolution images to arbitrary resolution, even up to ×60+ higher resolution.
- 22
- 22
- Recognizing pair-activities by causality analysis22
This article designs two sets of features for representing the pair-activities encoded as length-variable trajectory pairs, and develops a novel feature fusing procedure to learn the coefficients for weighting these features by maximizing the discriminating power measured by weighted correlation.
- 21
- Probabilistic Radiomics: Ambiguous Diagnosis with Controllable Shape Analysis20
This work proposes a controllable and explainable Probabilistic Radiomic framework, by combining the Radiomics analysis and probabilistic deep learning, and applies the proposed method on lung nodule diagnosis on LIDC-IDRI database to validate its effectiveness.
- 20
- Predicting Human Interaction via Relative Attention Model20
The results demonstrate that the proposed relative attention network successfully predicts informative regions between interacting subjects, which in turn yields superior human interaction prediction accuracy.
- Efficient region-aware large graph construction towards scalable multi-label propagation20
Extensive experiments on NUS-WIDE (260k images) and COREL-5k datasets validate the effectiveness and efficiency of the proposed framework for region-aware and scalable multi-label propagation.
- 19
- 19
- 19
- 18
- 18
- Frequency-Modulated Point Cloud Rendering with Easy Editing18
An effective point cloud rendering pipeline for novel view synthesis, which enables high fidelity local detail reconstruction, real-time rendering and user-friendly editing and high-fidelity interactive editing based on point cloud manipulation is developed.
- 18
- Mipmap-GS: Let Gaussians Deform with Scale-Specific Mipmap for Anti-Aliasing Rendering17
A unified optimization method to make Gaussians adaptive for arbitrary scales by self-adjusting the primitive properties and distribution by design pseudo ground-truth for the target scale and propose a scale-consistency guidance loss to inject scale information into 3D Gaussians.
- 17
- Movienet-PS: A Large-Scale Person Search Dataset in the Wild16
This paper introduces a new PS dataset, namely MovieNet-PS, based on an existing large-scale movie dataset, and proposes a novel global-local context network, which exploits scene and group context to boost the search performance.
- 16
- A distance measure between labeled combinatorial maps16
Experimental results show that the proposed inexact map matching approach produces richer search results than the exact map matching technique by tolerating small difference between maps.
- Searching Towards Class-Aware Generators for Conditional Generative Adversarial Networks15
This paper presents an idea that adopts neural architecture search to find a class-aware architecture for each class, and finds two intriguing phenomena that are inspirational to craft cGANs by hand.
- CooGAN: A Memory-Efficient Framework for High-Resolution Facial Attribute Editing15
A NOVEL pixel translation framework called Cooperative GAN(CooGAN) for HR facial image editing, which features a lighter selective transfer unit for more efficient multi-scale features fusion, yielding higher fidelity facial attributes manipulation.
- 14
- Anisotropic Stroke Control for Multiple Artists Style Transfer14
This work designs an Anisotropic Stroke Module (ASM) which realizes the dynamic adjustment of style-stroke between the non-trivial and the trivial regions and presents an novel Multi-Scale Projection Discriminator to realize the texture-level conditional generation.
- Fine-Grained Recognition via Attribute-Guided Attentive Feature Aggregation14
This work develops a novel attribute-guided attentive network to sequentially discover informative parts/regions, by seeking a good registration between attentive regions and predefined object attributes, which might be more robust than conventional attention model.
- Auto-generation of professional background music for home-made videos14
A joint probabilistic framework is proposed to model prior knowledge from two aspects, namely, the correlation between video and audio tracks, as well as the transition mode horn speech and music components for background music auto-generation.
- 13
- 13
- Skeleton2Humanoid: Animating Simulated Characters for Physically-plausible Motion In-betweening13
Experiments on the challenging LaFAN1 dataset show the proposed Skeleton2Humanoid system can outperform prior methods significantly in terms of both physical plausibility and accuracy.
- Pedestrian Detection via Bi-directional Multi-scale Analysis13
This work proposes a novel scale-context modeling scheme by exploiting the highly nonlinear dependency among scales by aggregating output response maps from mid-results of convolutional layers via a bi-directional recurrent sub-network.
- 13
- AMR-Transformer: Enabling Efficient Long-range Interaction for Complex Neural Fluid Simulation12
AMR-Transformer is proposed, an efficient and accurate neural CFD-solving pipeline that integrates a novel adaptive mesh refinement scheme with a Navier-Stokes constraint-aware fast pruning module that facilitates the modeling of global fluid wave patterns, such as turbulence and shockwaves.
- Intrinsic Phase-Preserving Networks for Depth Super Resolution12
A novel Phase-Preserving Filtering Module (PPFM) is developed to generate dynamic phase-aware filters according to the LR depth flow to filter out erroneous noisy components contained in RGB and then conduct depth enhancement via the modulation of the phase-preserved RGB signal.
- 12
- 12
- Object Wake-Up: 3D Object Rigging from a Single Image11
This work proposes an automated approach to build such 3D generic objects from single images and embed articulated skeletons in them and develops a novel skeleton prediction method with a multi-head structure for skeleton probability field estimation by utilizing the deep implicit functions.
- Hierarchical Classification of Pulmonary Lesions: A Large-Scale Radio-Pathomics Study11
This study solves the label ambiguity issue via a large-scale radio-pathomics dataset containing 5,134 radiological CT images with pathologically confirmed labels, and develops a three-level hierarchical classification system for pulmonary lesions, which covers most diseases in cancer-related diagnosis.
- 11
- Reinventing 2D Convolutions for 3D Medical Images11
This study proposes ACS (axial-coronal-sagittal) convolutions to perform natively 3D representation learning, while utilizing the pretrained weights from 2D counterparts, to bridge the gap between 2D and 3D convolution by reinventing the 2D convolutions.
- Flexible Network Binarization with Layer-Wise Priority11
A novel and flexible neural network binarization method is proposed by introducing the concept of layer-wise priority which binarizes parameters in inverse order of their layer depth by minimizing the discrepancy between the original real-valued weights and its binary approximations following block coordinate descent scheme.
- 11
- User Studies of a Multiplayer First Person Shooting Game with Tangible and Physical Interaction11
This system provides an intuitive way for the users to interact with the virtual world by physically moving around the real world and aiming freely with tangible objects to encourage physical interaction between the players as they compete or collaborate with each other.
- 10
- MIA-Prognosis: A Deep Learning Framework to Predict Therapy Response10
The proposed MIA-Prognosis framework with Measurement, Intervention and Assessment (MIA) information to predict therapy response achieves promising performance on predicting the immunotherapy response, and could further stratify low-risk and high-risk patients in terms of long-term survival.
- Fine-grained image analysis via progressive feature learning10
In this work, a two-stage framework for fine-grained recognition is developed that aims to discover and align local discriminative features, and the second stage aims to aggregate these features to get classification results.
- Human Motion Generation via Cross-Space Constrained Sampling10
A cross-space human motion video generation network which features two paths: a forward path that first samples/generates a sequence of low dimensional motion vectors based on Gaussian Process, and a backward path based on the predicted human images to re-extract the corresponding latent motion representations.
- Touch saliency10
This work proposes a new concept of touch saliency, and attempts to answer the question of whether the underlying image saliency map may be implicitly derived from the accumulative touch behaviors when many users browse the image on smart mobile devices with multi-touch display of small size.
- 3D surveillance system using multiple cameras10
This application is concerned with identifying humans in the scene and then identifying their postures, which can help with automatic threat assessment of a scene.
- Image Re-Emotionalizing9
A learning framework which discovers emotion-related knowledge, such as image local appearance distributions, from a set of emotion annotated images is proposed, which can yield effective and natural effects of the image affection generation process.
- 9
- Fast Fluid Simulation via Dynamic Multi-Scale Gridding8
This work hierarchically generate multi-scale micelles in Euclidean space by grouping particles that share similar motion patterns/characteristics based on super-light motion and scale estimation modules and achieves great simulation acceleration with negligible prediction error increase.
- Genopathomic profiling identifies signatures for immunotherapy response of lung adenocarcinoma via confounder-aware representation learning8
PITER was found to be a useful biomarker to identify patients of lung adenocarcinoma with both favorable progression-free and overall survival in the immunotherapy cohort, and indicated the biological basis in identifying patients with favorable outcome of immunotherapy.
- Progressive Stage-wise Learning for Unsupervised Feature Representation Enhancement8
This work explores new dimensions of unsupervised learning by proposing the Progressive Stage-wise Learning (PSL) framework, and discovers that by progressive stage-wise learning, un supervised feature representation can be effectively enhanced.
- CartoonRenderer: An Instance-Based Multi-style Cartoon Image Translator8
Extensive experimental results show the proposed novel "CartoonRenderer" framework produces higher quality cartoon style images than prior arts, with accurate semantic content preservation, and could easily process higher resolution photos, which is intractable for existing methods.
- Annotation modification for fine-grained visual recognition8
This work proposes a novel approach from a new viewpoint to solve the fine-grained recognition problem, namely annotation modification, and obtains some most confusing classes of each image through an easy-to-evaluate classifier, and modifications of each annotation using the active set of annotations.
- 8
- 7
- 7
- Representation-Agnostic Shape Fields7
Extensive experiments on diverse 3D representation formats, networks and applications, validate the universal effectiveness of the proposed Representation-Agnostic Shape Fields (RASF), a generalizable and computation-efficient shape embedding module for 3D deep learning.
- 7
- Human Action Transfer Based on 3D Model Reconstruction7
This paper designs a skeleton-to-3D-mesh generator to generate the 3D model, which achieves huge improvement on appearance reconstruction and adds a temporal connection to improve the smoothness of the model.
- Deep Cross-Modality Alignment for Multi-Shot Person Re-IDentification7
A deep cross-modality alignment network is proposed, which jointly explores both human sequence pairs and image pairs to facilitate training better multi-shot human Re-ID models, i.e., via transferring knowledge from image data to sequence data.
- 7
- Inferring Fluid Dynamics via Inverse Rendering6
This work achieves a photo-to-fluid-dynamics reconstruction functionality learned from unannotated videos, without any supervision of ground-truth fluid dynamics, by endowing each sampled point with a fluid volume value and derive a NeRF-like differentiable renderer dedicated from fluid data.
- 6
- BEST: Benchmark and Evaluation of Surveillance Task6
Compared with the existing surveillance/non-surveillance datasets, the proposed BEST dataset provides a realistic, extensive and diversified testbed for a more comprehensive performance evaluation, and experimental results show that, performance of seven pedestrian detection algorithms on BEST is worse than that on the existing datasets.
- 6
- 6
- 6
- 5
- 5
- USR: Unsupervised Separated 3D Garment and Human Reconstruction via Geometry and Semantic Consistency5
This work proposes an unsupervised separated 3D garments and human reconstruction model (USR), which reconstructs the human body and authentic textured clothes in layers without 3D models, and introduces SMPL-D model to show the benefit of the separated modeling of clothes and the humanBody that allows swapping clothes and virtual try-on.
- 5
- ViSTA: A Novel Network Improving Lung Adenocarcinoma Invasiveness Prediction from Follow-Up CT Series5
The proposed ViSTA framework is capable of improving performance close to the human level in the prediction of invasiveness of lung adenocarcinoma while being transferrable to other tasks analyzing serial medical data.
- 5
- Progressive Multi-granularity Analysis for Video Prediction5
This work proposes a progressive multiple granularity analysis framework that essentially decomposes the hard task into a series of more approachable sub-tasks where a better solution is easier to be sought and performs favorably against state-of-the-art prediction methods.
- 5
- Cross-modality motion parameterization for fine-grained video prediction5
This work proposes a novel cross-modality video generation network that directly links sound with the movement parameters of the operated object and automatically outputs corresponding object motion according to the rhythm of the given audio signal.
- Structure Guided Photorealistic Style Transfer5
This work designs a novel patch matching algorithm which simultaneously takes high-level category information and geometric structure information into account and achieves remarkable style transfer results on contents with geometric structure, including human body, vehicles, buildings, etc.
- Depth Structure Preserving Scene Image Generation5
A novel depth structure preserving scene image generation network (DSP-GAN), which favors a hierarchical architecture, which is built upon a Hawkes point process that models high-order spatial dependency between different depth layers.
- GeoMM: On Geodesic Perspective for Multi-modal Learning4
This paper introduces geodesic distance as a novel distance metric in multi-modal learning for the first time, to mine correlations between samples, aiming to address the limitations of common distance metric.
- 4
- Edit3D: Elevating 3D Scene Editing with Attention-Driven Multi-Turn Interactivity4
This work proposes a multi-turn editing framework and introduces an attention-driven open-set segmentation (ADSS) technique within this framework, which allows for more precise segmentation of parts, which enhances the editing precision and minimizes interference with pixels in areas that are not being edited.
- 4
- 4
- 4
- Weakly-supervised scene parsing with multiple contextual cues4
A weakly-supervised scene parsing algorithm that semantically parses a collection of images with multi-label, which is guided by the top-down category models and bottom-up local patch contexts across images that closely related segments usually have similar labels.
- Articulated Object Registration Using Simulated Physical Force/Moment for 3D Human Motion Tracking4
This paper presents a 3D registration algorithm based on simulated physical force/moment for articulated human motion tracking and introduces a hierarchical scheme for model state updating, which automatically incorporates human kinematic constraints.
- MSSIDD: A Benchmark for Multi-Sensor Denoising3
This paper introduces a new benchmark, the Multi-Sensor SIDD (MSSIDD) dataset, which is the first raw-domain dataset designed to evaluate the sensor transferability of denoising models, and proposes a sensor consistency training framework that enables denoising models to learn the sensor-invariant features, thereby facilitating the generalization of the consistent model to unseen sensors.
- Towards Interpreting and Utilizing Symmetry Property in Adversarial Examples3
This paper empirically proves this phenomenon is widespread and then analyzes the reason behind the existence of symmetry property in adversarial scenario by viewing adversarial attack in a fine-grained manner.
- 3
- SieveNet: Selecting Point-Based Features for Mesh Networks3
SieveNet is proposed, a novel paradigm that takes into account both the regular topology and the exact geometry of mesh neural networks and eliminates the need for hand-crafted feature engineering and can leverage off-the-shelf network architectures such as the vision transformer.
- HIRL: A General Framework for Hierarchical Image Representation Learning3
A general framework for Hierarchical Image Representation Learning (HIRL) is proposed, which aims to learn multiple semantic representations for each image, and these representations are structured to encode image semantics from fine-grained to coarse- grained.
- 3
- 3
- 3
- 3
- Pedestrian Detection Using ACF Based Fast R-CNN3
This paper proposes a novel detection framework named as ACF Based Fast R-CNN (ABF-CNN) and proposes to integrate Online Hard Example Mining (OHEM) training strategy into the authors' detection framework to solve the problem that CNN based methods have difficulty in hard negative mining.
- A unified supervised codebook learning framework for classification3
A discriminative visual dictionary learning method for boosting the classification performance by relaxing the hard binary constraints to soft nonnegative ones and a multiplicative nonnegative update procedure is proposed to optimize the objective function along with theoretic convergence proof.
- Vectorized Video Representation with Easy Editing via Hierarchical Spatio-Temporally Consistent Proxy Embedding2
This work proposes spatio-temporally consistent proxy nodes to represent dynamically changing objects/scenes in the video and achieves high video reconstruction accuracy with fewer parameters and supports complex video processing tasks, including video in-painting and keyframe-based temporally consistent video editing.
- Neural Block Compression: Variable Bitrates Feature Blocks for Texture Representation2
A novel multi-resolution texture compression scheme, Neural Block Compression (NBC), developed within the neural feature domain, which allocates higher bitrates to higher feature mip-levels and lower bitrates to lower feature mip-levels, thereby extending the concept of block compression from color domain into neural feature domain.
- 2
- 2
- 2
- Learning Unsupervised Gaze Representation via Eye Mask Driven Information Bottleneck2
A novel unsupervised/self-supervised gaze pre-training framework is proposed, which forces the full-face branch to learn a low dimensional gaze embedding without gaze annotations, through collaborative feature contrast and squeeze modules.
- 2
- RainNet: A Large-Scale Dataset for Spatial Precipitation Downscaling.2
This work presents the first \textbf{REAL} (non-simulated) Large-Scale Spatial Precipitation Downscaling Dataset, which contains 62,424 pairs of low-resolution and high-resolution precipitation maps for 17 years, and proposes an implicit physical estimation framework to learn the above characteristics.
- 2
- An efficient stochastic framework for 3D human motion tracking2
The proposed stochastic framework for articulated 3D human motion tracking uses explicit 3D reconstructions of the human body based on a visual hull algorithm as its system input, which effectively eliminates self-occlusion.
- Revisit Regularization Techniques for Gaze Estimation Generalization1
This paper finds gaze estimation is suffered from the over-fitting problem, and proposes to alleviate the problem with the technique of regularization: a) Weight Decay, b) Feature Normalization, and c) Weight Smoothing.
- 1
- 1
- Adversarial Attack on Video Retrieval1
This paper is the first to investigate adversarial examples on the video retrieval system in both non-targeted and targeted attack terms for copyright protection, and a triplet scheme is developed to take query-relevant and query-target pair-wise relationships together to enhance the attack performance.
- 1
- 1
- Real-Time Multiple Pedestrians Tracking in Multi-camera System1
A novel 3D realtime multi-view multi-target tracking framework featured by a global- to-local tracking by detection model and a 3D tracklet-to-tracklet data association scheme, compared to the state-of-the art methods.
- 1
- 1
- Resolution Booster: Global Structure Preserving Stitching Method for Ultra-High Resolution Image Translation1
Experimental results show the both capability of processing much higher resolution images while not decreasing the generating quality compared with baseline methods and generality of the model for flexible deployment.
- 1
- Long term activity prediction in first person viewpoint1
An asynchronous gaze-event driven activity evolving curve prediction network that generates event-modulated activity occurring probability functions via deriving long-temporal dependency from various events and extensive experimental results demonstrate the effectiveness of the proposed method in long-term activity evolving Curve prediction.
- Image Matching via Loopy RNN1
This work proposes a novel loopy recurrent neural network (Loopy RNN), which is capable of aggregating relationship information of two input images in a progressive/iterative manner and outputting the consolidated matching score in the final iteration.
- Human-Centric Images and Videos Analysis1
This tutorial reviews recent progresses in human-centric images and videos analysis: 1) fashion analysis: parsing, attribute prediction and retrieval; 2) action analysis: discriminative feature selection, pooling and fusion; and person verification: cross-domain person verification via learning a generalized similarity measure, and bit-scalable deep hashing with regularized similarity learning.
- 1
- Color grading method of stockings using color space and FCM clustering algorithm1
For the problem of inspecting the levelness of dyed polyamide filaments, sample images of stockings are extracted by CCD camera, and quotient image normalization method is proposed for the correction of optical source uniformity of the extracted stocking images, whose color feature processing is carried out by using the polynomial regression model.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Differentiable Projection from Optical Coherence Tomography B-Scan without Retinal Layer Segmentation Supervision–
This study presents a novel end-to-end framework to predict PMs from B-scans with differentiable operators, which produces high-quality PMs, significantly outperforming baselines, including a vanilla CNN without DPM and an optimization-based DPM without a deep prior.
- Gradient Correction beyond Gradient Descent–
A framework to perform gradient correction inspired by the idea of gradient prediction and based on Neural ODE, which can effectively improve the gradient quality to reduce training epochs by $\sim$ 20\% and also improve the network performance.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Multi-person/Group Interactive Video Generation–
This work proposes a novel human motion generation framework which can simultaneously consider the temporal coherence of each individual action, and consists of two components: Semantic Extractor, Motion Generator.
- –
- –
- –
- –
- Using machine vision to determine dyeing capability of nylon–
The experimental results show that the proposed method not only receives the quantification of the dyeing capability of nylon but also prevents the uneven dye and is effectively used in the practice of enterprise's production.
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.