Research map / Video / Action recognition
Learning Disentangled Classification and Localization Representations for Temporal Action Localization
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 |
|---|---|---|---|
| Disentangle proposal representations for classification and localization with D-Net (ablation) | A-Net baseline · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 53.50 -> 57.02 | Helped |
| Replace D-Net with one shared MLP (ablation) | A-Net baseline · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 57.02 -> 54.70 | Hurt |
| Replace D-Net with two different MLPs (ablation) | A-Net baseline · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 57.02 -> 55.09 | Hurt |
| Use disentangled features for their corresponding classification and localization tasks (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 47.16 -> 57.02 | Helped |
| Replace localization features with original proposal features for localization (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 53.72 -> 57.02 | Helped |
| Replace classification features with original proposal features for classification (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 56.76 -> 57.02 | Helped |
| Replace zero vectors with random vectors during feature decoding (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 57.02 -> 56.00 | Hurt |
| Use only common hidden variables (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 57.02 -> 54.26 | Hurt |
| Remove common hidden variables (ablation) | D-Net · THUMOS14 test | mAP at tIoU 0.5, RGB+Flow: 57.02 -> 55.75 | Hurt |
| Use disentangled classification and localization representations | Temporal action localization · ActivityNet v1.3 validation | average mAP at tIoU thresholds 0.5–0.95: 34.23 -> 35.24 | Helped |
| Use BMN proposals with the proposed method | Ours · ActivityNet v1.3 validation | average mAP at tIoU thresholds 0.5–0.95: 35.24 -> 38.45 | Helped |
Closest papers by meaning
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.91)
- Weakly Supervised Temporal Action Localization Through Learning Explicit Subspaces for Action and Context (AAAI 2021 · similarity 0.90)
- Test-Time Zero-Shot Temporal Action Localization (CVPR 2024 · similarity 0.88)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.88)
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.87)
- Prompt-guided Disentangled Representation for Action Recognition (NeurIPS 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.