Research map / Video / Video-language understanding
Asymmetric Cross-Guided Attention Network for Actor and Action Video Segmentation From Natural Language Query
Hao Wang, Cheng Deng, Junchi Yan, Dacheng Tao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add multi-resolution fusion (MRF) (ablation) | Baseline (RGB) · A2D Sentences | Overall IoU: 52.8 -> 57.7 | Improved |
| add multi-resolution fusion (MRF) (ablation) | Baseline (RGB) · A2D Sentences | Mean IoU: 44.1 -> 45.5 | Improved |
| add weighted binary cross entropy (WBCE) (ablation) | Baseline + MRF · A2D Sentences | Overall IoU: 57.7 -> 57.4 | No clear effect |
| add weighted binary cross entropy (WBCE) (ablation) | Baseline + MRF · A2D Sentences | Mean IoU: 45.5 -> 47.5 | Improved |
| add asymmetric cross-guided attention (ATT) (ablation) | Baseline + MRF + WBCE · A2D Sentences | Overall IoU: 57.4 -> 60.1 | Improved |
| add asymmetric cross-guided attention (ATT) (ablation) | Baseline + MRF + WBCE · A2D Sentences | mAP 0.5:0.95: 25.5 -> 27.4 | Improved |
| add asymmetric cross-guided attention (ATT) (ablation) | Baseline + MRF + WBCE · A2D Sentences | Mean IoU: 47.5 -> 49.0 | Improved |
| replace dynamic convolution with plain FCN on RGB input | Gavrilyuk et al. (RGB) · A2D Sentences | Overall IoU: 53.6 -> 52.8 | Hurt |
| fine-tune on A2D Sentences | Hu et al. · A2D Sentences | Overall IoU: 21.3 -> 47.4 | Improved |
| fine-tune on A2D Sentences | Li et al. · A2D Sentences | Overall IoU: 24.8 -> 51.5 | Improved |
| use RGB-only full model instead of RGB + Flow | Gavrilyuk et al. (RGB + Flow) · A2D Sentences | Overall IoU: 55.1 -> 60.1 | Improved |
| transfer to J-HMDB Sentences without fine-tuning | Ours (RGB) · J-HMDB Sentences | P@0.9: 0.0 | No clear effect |
Closest papers by meaning
- Actor and Action Video Segmentation From a Sentence (CVPR 2018 · similarity 0.91)
- Context Modulated Dynamic Networks for Actor and Action Video Segmentation with Language Queries (AAAI 2020 · similarity 0.91)
- Collaborative Spatial-Temporal Modeling for Language-Queried Video Actor Segmentation (CVPR 2021 · similarity 0.90)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.88)
- Kronecker Mask and Interpretive Prompts are Language-Action Video Learners (ICLR 2025 · similarity 0.87)
- ResCLIP: Residual Attention for Training-free Dense Vision-language Inference (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
