Research map / Video / Action recognition
Fine-grained Video Categorization with Redundancy Reduction Attention
Chen Zhu, Xiao Tan, Feng Zhou, Xiao Liu, Kaiyu Yue, Errui Ding, Yi Ma
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 L_e (ensemble loss) to L_i (ablation) | RRA · ActivityNet v1.3 val | mAP (ensemble): 82.97 -> 83.42 | Improved |
| add L_c (concatenation loss) to L_e+L_i (ablation) | RRA · ActivityNet v1.3 val | mAP (ensemble): 83.42 -> 82.59 | Hurt |
| add L_c to L_i (ablation) | RRA · ActivityNet v1.3 val | mAP (ensemble): 82.97 -> 82.80 | Hurt |
| train with L_e alone (no per-glimpse supervision) (ablation) | RRA · ActivityNet v1.3 val | mAP (ensemble): 83.42 -> 36.24 | Hurt |
| increase glimpses from 1 to 4 (ablation) | RRA · ActivityNet v1.3 val | mAP: 80.89 -> 83.42 | Improved |
| increase glimpses from 4 to 5 (ablation) | RRA · ActivityNet v1.3 val | mAP: 83.42 -> 82.94 | Hurt |
| add spatio-temporal attention instead of average pooling (ablation) | RRA · ActivityNet v1.3 val | mAP: 80.20 -> 83.42 | Improved |
| use spatio-temporal attention instead of spatial attention with temporal average pooling (ablation) | RRA · ActivityNet v1.3 val | mAP: 81.97 -> 83.42 | Improved |
| add batch normalization in the reduction module (ablation) | RRA · ActivityNet v1.3 val | mAP: 82.41 -> 83.42 | Improved |
| add ReLU in the reduction module (ablation) | RRA · ActivityNet v1.3 val | mAP: 83.15 -> 83.42 | Improved |
| use tanh activation rather than none or -ReLU(x) (ablation) | RRA · ActivityNet v1.3 val | mAP: 82.75 -> 83.42 | Improved |
| use sequential redundancy-reduction glimpses instead of 4 parallel glimpses (ablation) | RRA · ActivityNet v1.3 val | mAP: 82.39 -> 83.42 | Improved |
Closest papers by meaning
- Attention Convolutional Binary Neural Tree for Fine-Grained Visual Categorization (CVPR 2020 · similarity 0.85)
- Learning Fine-Grained Representations through Textual Token Disentanglement in Composed Video Retrieval (ICLR 2025 · similarity 0.84)
- Purely Attention Based Local Feature Integration for Video Classification (TPAMI 2020 · similarity 0.84)
- Data-free Knowledge Distillation for Fine-grained Visual Categorization (ICCV 2023 · similarity 0.84)
- Attention Clusters: Purely Attention Based Local Feature Integration for Video Classification (CVPR 2018 · similarity 0.84)
- HyperVAttention: Efficient Sparse Attention with Spatio-Temporal Clustering for Video Diffusion (NeurIPS 2026 · similarity 0.82)
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.
