Research map / Architectures and efficiency / RNNs and normalization
Ultrafast Video Attention Prediction with Coupled Knowledge Distillation
Kui Fu, Peipei Shi, Yafei Song, Shiming Ge, Xiangju Lu, Jia Li
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 |
|---|---|---|---|
| reduce input resolution from 256x256 to 64x64 (ablation) | UVA-Net (DVA teacher) · AVS1K | NSS: 1.447 -> 1.981 | Improved |
| reduce input resolution further from 64x64 to 32x32 (ablation) | UVA-Net (DVA teacher) · AVS1K | NSS: 1.981 -> 1.905 | Hurt |
| use DVA instead of SalNet as spatial teacher (ablation) | UVA-Net 64x64 · AVS1K | NSS: 1.955 -> 1.981 | Improved |
| add coupled knowledge distillation (MobileNetV2 blocks) (ablation) | UVA-Net MB · AVS1K | NSS: 1.906 -> 1.943 | Improved |
| add coupled knowledge distillation (CA-Res blocks) (ablation) | UVA-Net CA-Res · AVS1K | NSS: 1.921 -> 1.981 | Improved |
| replace MobileNetV2 block with CA-Res channel attention block (with distillation) (ablation) | UVA-Net · AVS1K | NSS: 1.943 -> 1.981 | Improved |
| replace SE block with CA-Res channel attention (avg+max pooling) (ablation) | UVA-Net with distillation · AVS1K | NSS: 1.943 -> 1.981 | Improved |
| add spatiotemporal joint optimization fusing temporal branch (vs spatial branch alone) (ablation) | UVA-Net CA-Res+dis · AVS1K | NSS: 1.864 -> 1.981 | Improved |
| add spatiotemporal joint optimization fusing spatial branch (vs temporal branch alone) (ablation) | UVA-Net CA-Res+dis · AVS1K | NSS: 1.783 -> 1.981 | Improved |
| distill from DVA into 64x64 student vs using DVA directly | UVA-DVA-64 vs DVA · AVS1K | NSS: 2.044 -> 1.981 | Hurt |
Closest papers by meaning
- Recurrent Mixture Density Network for Spatiotemporal Visual Attention (ICLR 2017 · similarity 0.87)
- VideoNSA: Native Sparse Attention Scales Video Understanding (ICLR 2026 · similarity 0.87)
- AttentionNAS: Spatiotemporal Attention Cell Search for Video Classification (ECCV 2020 · similarity 0.86)
- A^2-Nets: Double Attention Networks (NeurIPS 2018 · similarity 0.86)
- Faster Video Diffusion with Trainable Sparse Attention (NeurIPS 2025 · similarity 0.85)
- Bidirectional Sparse Attention for Faster Video Diffusion Training (NeurIPS 2026 · similarity 0.85)
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.
