Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks80 from public data
- Rethinking BiSeNet For Real-time Semantic Segmentation784
A novel and efficient structure named Short-Term Dense Concatenate network (STDC network) is proposed by removing structure redundancy by gradually reducing the dimension of feature maps and use the aggregation of them for image representation, which forms the basic module of STDC network.
- 606
- Large Scale Visual Food Recognition255
This paper introduces Food2K, which is the largest food recognition dataset with 2,000 categories and over 1 million images, and proposes a deep progressive region enhancement network for food recognition, which mainly consists of two components, namely progressive local feature learning and region feature enhancement.
- Accurate realtime full-body motion capture using a single depth camera246
The idea is to formulate the registration problem in a Maximum A Posteriori (MAP) framework and iteratively register a 3D articulated human body model with monocular depth cues via linear system solvers.
- PromptDet: Towards Open-Vocabulary Detection Using Uncurated Images245
The goal of this work is to establish a scalable pipeline for expanding an object detector towards novel/unseen categories, using zero manual annotations, and the proposed detector, termed as PromptDet, shows superior performance over existing approaches with fewer additional training images andzero manual annotations whatsoever.
- 173
- ISIA Food-500: A Dataset for Large-Scale Food Recognition via Stacked Global-Local Attention Network162
This work introduces the dataset ISIA Food-500 with 500 categories from the list in the Wikipedia and 399,726 images, a more comprehensive food dataset that surpasses existing popular benchmark datasets by category coverage and data volume, and proposes a stacked global-local attention network, which consists of two sub-networks for food recognition.
- 157
- Towards Accurate Post-Training Quantization for Vision Transformer136
This work proposes a novel Accurate Post-training Quantization framework for Vision Transformer, namely APQ-ViT, which surpasses the existing post-training quantization methods by convincing margins, especially in lower bit-width settings.
- MobileVLM : A Fast, Strong and Open Vision Language Assistant for Mobile Devices129
This work presents MobileVLM, a competent multimodal vision language model (MMVLM), an amalgamation of a myriad of architectural designs and techniques that are mobile-oriented, which comprises a set of language models at the scale of 1.4B and 2.7B trained from scratch and demonstrates on par performance compared with a few much larger models.
- VideoMocap123
The power and effectiveness of the video-based motion modeling process in an image-based keyframe animation framework is demonstrated by generating a wide variety of physically realistic human actions from uncalibrated monocular video sequences such as sports video footage.
- Realtime human motion control with a small number of inertial sensors110
This paper introduces an approach to performance animation that employs a small number of motion sensors to create an easy-to-use system for an interactive control of a full-body human character.
- Structure Guided Lane Detection100
A top-down vanishing point guided anchoring mechanism is proposed to produce intensive anchors, which efficiently capture various lanes, and multi-level structural constraints are used to improve the perception of lanes.
- 86
- 85
- Trash to Treasure: Harvesting OOD Data with Cross-Modal Matching for Open-Set Semi-Supervised Learning81
This work proposes a novel cross-modal matching strategy to detect OOD samples that substantially lifts the performance on open-set SSL and outperforms the state-of-the-art by a large margin.
- 76
- Physically valid statistical models for human motion generation67
This article shows how statistical motion priors can be combined seamlessly with physical constraints for human motion modeling and generation and effectively utilize the new model to generate a wide range of natural-looking motions that achieve the goals specified by users.
- 56
- 56
- 52
- 46
- Ingredient-Guided Region Discovery and Relationship Modeling for Food Category-Ingredient Prediction41
- Weakly Supervised Semantic Segmentation Via Progressive Patch Learning37
A novel “Progressive Patch Learning” approach is proposed to improve the local details extraction of the classification, producing the CAM better covering the whole object rather than only the most discriminative regions as in CAMs obtained in conventional classification models.
- Two-stage Visual Cues Enhancement Network for Referring Image Segmentation31
Through the two-stage enhancement, the proposed TV-Net enjoys better performances in learning fine-grained matching behaviors between the natural language expression and image, especially when the visual information of the referent is inadequate, thus produces better segmentation results.
- 28
- 19
- 19
- 18
- Concept Propagation via Attentional Knowledge Graph Reasoning for Video-Text Retrieval18
A simple yet effective two-stream framework which takes the concept information into account and introduces a new branch of semantic-level matching, and a concept propagation mechanism for mining the latent semantics in videos and achieving enriched representations.
- 16
- PPMN: Pixel-Phrase Matching Network for One-Stage Panoptic Narrative Grounding16
A one-stage end-to-end Pixel-Phrase Matching Network (PPMN), which directly matches each phrase to its corresponding pixels instead of region proposals and outputs panoptic segmentation by simple combination, which achieves new state-of-the-art performance on the PNG benchmark with 4.0 absolute Average Recall gains.
- Zero-shot Video Classification with Appropriate Web and Task Knowledge Transfer14
A model is constructed based on the constructed attributes, the mined category-object relationships and the proposed Task Similarity aware Representation Learning (TSRL) to encourage a more reliable transfer of knowledge transfer in zero-shot video classification.
- Rethinking the Optimization of Average Precision: Only Penalizing Negative Instances before Positive Ones Is Enough13
This work proposes a novel loss, namely Penalizing Negative instances before Positive ones (PNP), which can directly minimize the number of negative instances before each positive one and systematically investigates different gradient assignment solutions via constructing derivative functions of the loss.
- 13
- 13
- 12
- 12
- 11
- 11
- 10
- Synthesizing Counterfactual Samples for Effective Image-Text Matching10
Compared with traditional hard negative mining, the proposed Counterfactual Matching framework largely alleviates the over-fitting phenomenon and effectively captures the fine-grained correlations between image and text modality.
- 10
- Free-Form Image Inpainting via Contrastive Attention Network9
A self-supervised Siamese inference network that can encode contextual semantics from full resolution images and obtain more discriminative representations and a multi-scale decoder with a novel dual attention fusion module (DAF), which can combine both the restored and known regions in a smooth way.
- Interactive Tracking of 2D Generic Objects with Spacetime Optimization9
The problem is formulated in a spacetime optimization framework that optimizes over the whole sequence simultaneously, and the resulting solution is consistent with visual measurements across the entire sequence while satisfying user constraints.
- Enriching Phrases with Coupled Pixel and Object Contexts for Panoptic Narrative Grounding8
This work proposes to enrich phrases with coupled pixel and object contexts by designing a Phrase-Pixel-Object Transformer Decoder (PPO-TD), where both fine-grained part details and coarse-graining entity clues are aggregated to phrase features.
- MT-Net Submission to the Waymo 3D Detection Leaderboard8
This work designs a 2D backbone to utilize multi-scale features for better detecting objects with various sizes, together with an optimal transport-based target assignment strategy, which dynamically assigns richer supervision signals to the detection candidates.
- 8
- 8
- Low-Dose CT Image Super-resolution Network with Noise Inhibition Based on Feedback Feature Distillation Mechanism7
The proposed LDCT joint SR and denoising reconstruction network has been extensively evaluated through ablation, quantitative, and qualitative experiments and demonstrates that the method can recover noise-free and detail-sharp images, resulting in better reconstruction results.
- 7
- Exploration and Exploitation of Unlabeled Data for Open-Set Semi-supervised Learning6
This paper proposes a prototype-based clustering and identification algorithm that explores the inherent similarity and difference among samples at feature level and effectively cluster them around several predefined ID and OOD prototypes, thereby enhancing feature learning and facilitating ID/OOD identification.
- 6
- 5
- ROME: Robustifying Memory-Efficient NAS via Topology Disentanglement and Gradients Accumulation5
A new algorithm called RObustifying Memory-Efficient NAS (ROME), which involves separate parameters to disentangle the topology from the operations of the architecture and enforce fair sampling for weight update and apply a gradient accumulation mechanism for architecture parameters.
- 5
- An Improved Convolutional Block Attention Module for Chinese Character Recognition4
A novel method named ICBAM (Improved Convolutional Block Attention Module) for Chinese character recognition in the wild is proposed, which improves the generalization performance of the network and effectively avoid over-fitting.
- 4
- 3
- Meta-Ensemble Parameter Learning3
WeightFormer is introduced, a Transformer-based model that can predict student network weights layer by layer in a forward pass, according to the teacher model parameters, which can potentially lead to a new, more efficient, and scalable paradigm of ensemble networks parameter learning.
- Development and validation of multiple machine learning algorithms for the classification of G-protein-coupled receptors using molecular evolution model-based feature extraction strategy3
The potential of the novel feature extraction strategy is demonstrated and an effective theoretical design for the hierarchical classification of GPCRs is provided and multiple machine learning models based on Naïve Bayesian, random forest, support vector machine, and multi-layer perceptron algorithms are created.
- 2
- 3D Colored Shape Reconstruction from a Single RGB Image through Diffusion2
Experimental results demonstrate that the proposed method achieves competitive performance on colored 3D shape reconstruction, and the ablation study validates the positive role of the color prediction module in improving the reconstruction quality of 3D geometric point cloud.
- 2
- ReADS: A Rectified Attentional Double Supervised Network for Scene Text Recognition2
A Rectified Attentional Double Supervised Network (ReADS) for general scene text recognition is designed with different modules in two supervised branches which can make a complementary to each other to overcome the weakness of CTC and Attn.
- 2
- 2
- 2
- 3rd Place Solution for PVUW Challenge 2023: Video Panoptic Segmentation1
This work regards the video panoptic segmentation task as a segmentation target querying task, represent both semantic and instance targets as a set of queries, and then combine these queries with video features extracted by neural networks to predict segmentation masks.
- 1
- 1
- ALEC: An Accurate, Light and Efficient Network for CAPTCHA Recognition1
An Accurate, Light and Efficient network for CAPTCHA recognition (ALEC) based on the encoder-decoder structure is proposed, which can greatly reduce the computation complexity and parameters while ensuring the recognition accuracy.
- –
- –
- –
- Human Emergency Behaviour Characteristics in Large-scale Emergencies Based on Mobile Communication Data–
It is found that the communication peaks accompanying an emergency are local in time, resulting in a communication avalanche that importantly engages eyewitness social networks and has potential application value in predicting the impact degree of crises and the analysis and classification of human social attributes.
- A Method for Scene Text Style Transfer–
This work proposes a novel end-to-end scene text style transfer framework that simultaneously translates the text instance and scene background with different styles and demonstrates that the synthetic images generated by the model can benefit the scene text detection task.
- –
- –
- –
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.