Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks129 from public data
- 531
- RMT: Retentive Networks Meet Vision Transformers270
This work extends the RetNet's temporal decay mechanism to the spatial do-main, and proposes a spatial decay matrix based on the Manhattan distance to introduce the explicit spatial prior to Self-Attention, aiming to reduce the computational burden of modeling global information without disrupting the spa-tial decay matrix.
- 151
- 148
- 124
- Vision Transformer with Super Token Sampling122
A simple yet strong super token attention (STA) mechanism with three steps: the first samples super tokens from visual tokens via sparse association learning, the second performs self-attention on super tokens, and the last maps them back to the original token space.
- Disentangled Variational Representation for Heterogeneous Face Recognition97
This paper model a face representation with an intrinsic identity information and its within-person variations and proposes an alternative optimization scheme for the disentangled variational representation part and the heterogeneous face recognition network part that effectively reduces the NIR and VIS domain discrepancy and alleviates over-fitting.
- Wavelet Domain Generative Adversarial Network for Multi-scale Face Hallucination91
This paper proposes a wavelet-domain generative adversarial method that can ultra-resolve a very low-resolution face image and can significantly improve identification accuracy for low-resolution face images captured in the wild.
- MMFakeBench: A Mixed-Source Multimodal Misinformation Detection Benchmark for LVLMs83
MMFakeBench is introduced, the first comprehensive benchmark for mixed-source MMD, and MMD-Agent is proposed, a novel approach to integrate the reasoning, action, and tool-use capabilities of LVLM agents, significantly enhancing accuracy and generalization.
- FKA-Owl: Advancing Multimodal Fake News Detection through Knowledge-Augmented LVLMs81
This paper proposes FKA-Owl, a novel framework that leverages forgery-specific knowledge to augment LVLMs, enabling them to reason about manipulations effectively and achieves superior cross-domain performance compared to previous methods.
- A Survey of Deep Facial Attribute Analysis72
A general pipeline that deep facial attribute analysis follows, which comprises two stages: data preprocessing and model construction is summarized and the underlying theories of this two-stage pipeline for both FAE and FAM are introduced.
- 71
- Rethinking Local Perception in Lightweight Vision Transformer66
CloFormer is introduced, a lightweight vision transformer that leverages context-aware local enhancement and introduces AttnConv, a convolution operator in attention's style that enables the model to perceive high-frequency and low-frequency information.
- Arbitrary Talking Face Generation via Attentional Audio-Visual Coherence Learning65
A novel arbitrary talking face generation framework is proposed by discovering the audio-visual coherence via the proposed Asymmetric Mutual Information Estimator (AMIE) and a Dynamic Attention (DA) block by selectively focusing the lip area of the input image during the training stage, to further enhance lip synchronization.
- 64
- Selective Wavelet Attention Learning for Single Image Deraining60
A selectiveWavelet attention learning method by learning a series of wavelet attention maps to guide the separation of rain and background information in both spatial and frequency domains that achieves more appealing results both quantitatively and qualitatively than the recent state-of-the-art methods.
- Breaking the Low-Rank Dilemma of Linear Attention54
Rank-Augmented Linear Attention (RALA), which rivals the performance of Softmax attention while maintaining linear complexity and high efficiency, is introduced and the Rank-Augmented Vision Linear Transformer (RAVLT) is constructed.
- 49
- 47
- 43
- Jailbreak Attacks and Defenses against Multimodal Generative Models: A Survey42
A detailed taxonomy of attack methods, defense mechanisms, and evaluation frameworks specific to multimodal generative models is presented, including modalities such as Any-to-Text, Any-to-Vision, and Any-to-Any within generative systems.
- LAMP-HQ: A Large-Scale Multi-pose High-Quality Database and Benchmark for NIR-VIS Face Recognition42
A benchmark is furnished along with the protocol for NIR-VIS face recognition via generation on LAMP-HQ and a spectral approach is proposed to reduce the domain performance to improve the performance of the model training process.
- Learning Disentangled Representation for Cross-Modal Retrieval with Deep Mutual Information Estimation40
Extensive experiments on two large-scale benchmark datasets demonstrate that the proposed method can learn better visual-semantic embedding and achieve state-of-the-art cross-modal retrieval results.
- Heterogeneous Test-Time Training for Multi-Modal Person Re-identification36
A Cross-identity Inter-modal Margin (CIM) loss is proposed to amplify the differentiation among distinct identity samples and a Multi-modal Test-time Training (MTT) strategy is designed to enhance the generalization of the model by leveraging the relationships in the heterogeneous modalities and the information existing in the test data.
- ViTAR: Vision Transformer with Any Resolution36
This work proposes a novel module for dynamic resolution adjustment, designed with a single Transformer block, specifically to achieve highly efficient incremental token integration and introduces fuzzy positional encoding in the Vision Transformer to provide consistent positional awareness across multiple resolutions, thereby preventing overfitting to any single training resolution.
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.