Research map / Video / Action recognition
Learning Salient Boundary Feature for Anchor-free Temporal Action Localization
Check 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 tIoU-trained quality confidence (ablation) | AFSD basic predictor · THUMOS14 | average mAP: 40.4 -> 41.4 | Helped |
| replace quality confidence with centerness (ablation) | AFSD basic predictor · THUMOS14 | mAP@0.7: 19.0 -> 17.7 | Hurt |
| use tanh normalization instead of hard clipping (ablation) | AFSD activation-guided learning · THUMOS14 | average mAP: 42.6 -> 43.5 | Helped |
| use tanh normalization instead of min-max normalization (ablation) | AFSD activation-guided learning · THUMOS14 | average mAP: 42.9 -> 43.5 | Helped |
| focus boundary pooling more on the action than the background (ablation) | AFSD boundary feature extraction · THUMOS14 | average mAP: 42.6 -> 43.5 | Helped |
| use only frame-level features rather than the naive convolutional refinement (ablation) | AFSD refinement · THUMOS14 | average mAP: 41.6 -> 40.5 | Hurt |
| add saliency-based boundary refinement to naive convolutional refinement (ablation) | AFSD · THUMOS14 | average mAP: 41.6 -> 43.5 | Helped |
| replace max boundary pooling with mean pooling (ablation) | AFSD boundary pooling · THUMOS14 | average mAP: 43.5 -> 42.7 | Hurt |
| replace max boundary pooling with temporal convolution (ablation) | AFSD boundary pooling · THUMOS14 | average mAP: 43.5 -> 42.6 | Hurt |
| replace max boundary pooling with feature stacking (ablation) | AFSD boundary pooling · THUMOS14 | average mAP: 43.5 -> 42.9 | Hurt |
| add activation-guided loss to boundary contrastive learning (ablation) | AFSD Boundary Consistency Learning · THUMOS14 | average mAP: 42.7 -> 43.5 | Helped |
| add boundary contrastive loss to activation-guided learning (ablation) | AFSD Boundary Consistency Learning · THUMOS14 | average mAP: 42.7 -> 43.5 | Helped |
Closest papers by meaning
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.91)
- Temporal Action Localization by Structured Maximal Sums (CVPR 2017 · similarity 0.90)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.90)
- Learning To Refactor Action and Co-Occurrence Features for Temporal Action Localization (CVPR 2022 · similarity 0.90)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.89)
- Boundary Denoising for Video Activity Localization (ICLR 2024 · similarity 0.89)
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.