Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks260 from public data
- 233
- Reasoning with Heterogeneous Graph Alignment for Video Question Answering225
This work proposes a deep heterogeneous graph alignment network over the video shots and question words and explores the network architecture from four steps: representation, fusion, alignment, and reasoning.
- 198
- Vector-Decomposed Disentanglement for Domain-Invariant Object Detection147
This paper tries to disentangle domain-invariant representations from domain-specific representations, and proposes a novel disentangled method based on vector decomposition.
- 139
- Curls & Whey: Boosting Black-Box Adversarial Attacks127
The Curls & Whey attack is proposed, which extends the attack to the targeted misclassification, effectively reducing the difficulty of targeted attacks under black-box condition and shows promising transferability against ensemble models as well as adversarially trained models.
- Bidirectional Adversarial Training for Semi-Supervised Domain Adaptation123
A Bidirectional Adversarial Training (BiAT) network is devised to perform diverse adversarial trainings jointly and evaluates the effectiveness of BiAT on three benchmark datasets and experimental results demonstrate the proposed method achieves the state-of-the-art.
- 121
- Universal-Prototype Enhancing for Few-Shot Object Detection120
A new framework of few-shot object detection with universal prototypes (F SODup) that owns the merit of feature generalization towards novel objects is developed and a new prototype is proposed that is learned from all object categories.
- Multi-modal Circulant Fusion for Video-to-Language and Backward118
A new approach of multi-modal fusion, namely Multi- modal Circulant Fusion (MCF), which defines two types of interaction operations between vectors and matrices and obtains the state-of-the-art for video captioning.
- 104
- Multi-label boosting for image annotation by structural grouping sparsity95
The proposed Multi-label Boosting by the selection of heterogeneous features with structural Grouping Sparsity (MtBGS) has better multi-label image annotation performance and leads to a quite interpretable model for image understanding.
- 91
- 86
- 81
- Understanding the effective receptive field in semantic image segmentation76
A Dense Global Context Module is designed, which makes the effective receptive field coverage larger and density higher and reduces a large number of parameters while the performance has been substantially improved in segmentation model.
- Catching the Temporal Regions-of-Interest for Video Captioning75
A novel approach which will automatically focus on regions-of-interest and catch their temporal structures and obtains the state-of theart results on popular evaluation metrics like BLEU-4, CIDEr, and METEOR.
- 71
- 69
- Weakly-Supervised Video Anomaly Detection With Snippet Anomalous Attention68
This paper proposes an Anomalous Attention mechanism for weakly-supervised anomaly detection that takes into account snippet-level encoded features without the supervision of pseudo labels, and obtains a more precise anomaly localization.
- 66
- Behavior-Contextualized Item Preference Modeling for Multi-Behavior Recommendation61
This paper introduces a novel approach, Behavior-Contextualized Item Preference Modeling (BCIPM), which discerns and learns users' specific item preferences within each behavior and considers only those preferences relevant to the target behavior for final recommendations, significantly reducing noise from auxiliary behaviors.
- Explore Multi-Step Reasoning in Video Question Answering58
This paper designs a system to automatically generate a large-scale dataset, namely SVQA (Synthetic Video Question Answering), and develops a new VideoQA model that contains a new attention module, which contains spatial attention mechanism to address crucial and multiple logical sub-tasks embedded in questions, as well as a refined GRU called ta-GRU (temporal-attention GRU).
- Adaptive iterative attack towards explainable adversarial robustness55
Ada-FGSM (Adaptive FGSM), a new iterative attack that adaptively allocates step size of noises according to gradient information at each step is proposed, proving that if the upper bound of noise added to the original image is fixed, the attack effect can be improved if the step size is positively correlated with the gradient obtained at each steps by querying the target model.
- 54
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 2026-10-13. 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.