Research map / Video / Action recognition
Weakly-Supervised Video Re-Localization with Multiscale Attention Model
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 multiscale attention module (dilated, varied kernel sizes) (ablation) | weakly supervised base model · Video re-localization, weakly supervised, test | avg top-1 mAP: 18.0 -> 19.2 | Helped |
| add self-attention features (ablation) | weakly supervised base model · Video re-localization, weakly supervised, test | avg top-1 mAP: 18.0 -> 19.7 | Helped |
| combine multiscale and self-attention (full model) (ablation) | weakly supervised base model · Video re-localization, weakly supervised, test | avg top-1 mAP: 18.0 -> 20.7 | Helped |
| add co-attention loss to weighted cross entropy (ablation) | fully supervised model · Video re-localization, fully supervised, test | avg top-1 mAP: 27.2 -> 28.2 | Helped |
| set co-attention loss ratio alpha to 0.3 instead of 1 (ablation) | weakly supervised model · Video re-localization, weakly supervised, test | avg top-1 mAP: 18.4 -> 20.7 | Helped |
| set co-attention loss ratio alpha to 0.3 instead of 0 (no co-attention term) (ablation) | weakly supervised model · Video re-localization, weakly supervised, test | avg top-1 mAP: 19.0 -> 20.7 | Helped |
| use model's attention modules with cross entropy only instead of CGBM | fully supervised model · Video re-localization, fully supervised, test | avg top-1 mAP: 25.7 -> 27.2 | Helped |
| use 1D conv with kernel size 1 in multiscale heads (ablation) | multiscale attention · weakly supervised | performance | Hurt |
Closest papers by meaning
- Weakly Supervised Video Moment Localization with Contrastive Negative Sample Mining (AAAI 2022 · similarity 0.88)
- TransVCL: Attention-Enhanced Video Copy Localization Network with Flexible Supervision (AAAI 2023 · similarity 0.88)
- Spatio-Temporal Video Re-Localization by Warp LSTM (CVPR 2019 · similarity 0.88)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.86)
- Transferable Video Moment Localization by Moment-Guided Query Prompting (AAAI 2024 · similarity 0.86)
- Omnipotent Distillation with LLMs for Weakly-Supervised Natural Language Video Localization: When Divergence Meets Consistency (AAAI 2024 · similarity 0.86)
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.