Research map / Video / Action recognition
Finding Action Tubes with a Sparse-to-Dense Framework
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 long-term feature augmentation (LFA) (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 62.2 -> 71.1 | Helped |
| add embedding vectors for box association (ablation) | Sparse-to-dense action-tube detector with LFA and shift vectors · UCF101-24 | v-mAP@0.3: 68.9 -> 71.1 | Helped |
| add shift vectors for box association (ablation) | Sparse-to-dense action-tube detector with LFA and embedding vectors · UCF101-24 | v-mAP@0.3: 69.6 -> 71.1 | Helped |
| replace fixed average point sampling with adaptive dynamic temporal sampling (DTS) (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 67.4 -> 71.1 | Helped |
| replace fixed 4-point sampling with DTS (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 63.3 -> 71.1 | Helped |
| replace fixed 10-point sampling with DTS (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 67.5 -> 71.1 | Helped |
| replace fixed average-step sampling with DTS (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 66.9 -> 71.1 | Helped |
| use fixed 4-point sampling instead of fixed 10-point sampling (ablation) | Sparse-to-dense action-tube detector · UCF101-24 | v-mAP@0.3: 67.5 -> 63.3 | Hurt |
| use the sparse-to-dense framework instead of the strongest listed prior method | Action-tube detection · JHMDB-21, Δ=0.2 | v-mAP: 82.3 -> 76.1 | Hurt |
| use the sparse-to-dense framework instead of the strongest listed prior method | Action-tube detection · JHMDB-21, Δ=0.5 | v-mAP: 80.5 -> 74.3 | Hurt |
| use the sparse-to-dense framework instead of the strongest listed prior method | Action-tube detection · UCFSports, Δ=0.2 | v-mAP: 97.8 -> 94.3 | Hurt |
| use the sparse-to-dense framework instead of the strongest listed prior method | Action-tube detection · UCFSports, Δ=0.5 | v-mAP: 97.8 -> 93.8 | Hurt |
Closest papers by meaning
- TubeR: Tubelet Transformer for Video Action Detection (CVPR 2022 · similarity 0.89)
- STEP: Spatio-Temporal Progressive Learning for Video Action Detection (CVPR 2019 · similarity 0.89)
- End-to-End Spatio-Temporal Action Localisation with Video Transformers (CVPR 2024 · similarity 0.87)
- Temporal Action Detection Model Compression by Progressive Block Drop (CVPR 2025 · similarity 0.85)
- Taylor Videos for Action Recognition (ICML 2024 · similarity 0.84)
- End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames (CVPR 2024 · similarity 0.84)
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.