Research map / Video / Action recognition
Tubelets: Unsupervised Action Proposals from Spatiotemporal Super-Voxels
Mihir Jain, Jan C. van Gemert, Hervé Jeǵou, Patrick Bouthémy, Cees G. M. Snoek
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 |
|---|---|---|---|
| segment iMotion maps instead of RGB video (ablation) | super-voxel segmentation · UCF Sports train | Recall: 17.3 -> 53.2 | Improved |
| add Motion similarity to grouping (All vs All-but-motion) (ablation) | super-voxel grouping, video segmentation · UCF Sports train | MABO: 53.4 -> 58.1 | Improved |
| use union of five grouping functions instead of best single (ablation) | super-voxel grouping, video segmentation · UCF Sports train | Recall: 66.7 -> 74.7 | Improved |
| combine Tubelets from video and iMotion segmentation (ablation) | Tubelets · UCF Sports train | Recall: 77.0 -> 93.6 | Improved |
| add motion pruning (ablation) | Tubelets · UCF Sports | Recall: 93.5 -> 93.5 | No clear effect |
| add overlap pruning (ablation) | Tubelets · UCF Sports | Recall: 93.5 -> 90.5 | Hurt |
| add spatial refinement after pruning (ablation) | Tubelets · UCF Sports | Recall: 90.5 -> 91.9 | Improved |
| add temporal refinement (ablation) | Tubelets · MSR-II | Recall: 30.1% in Recall and 9.3% in MABO | Improved |
| add spatial refinement after temporal refinement (ablation) | Tubelets · MSR-II | Recall: 35.2 -> 47.4 | Improved |
| add motion pruning (ablation) | Tubelets · MSR-II | Recall: 5.1 -> 5.1 | No clear effect |
| evaluate spatial-only instead of spatiotemporal overlap (ablation) | Tubelets · MSR-II | Recall: 2.2 -> 81.3 | Improved |
| add overlap pruning (ablation) | Tubelets · UCF101 | Recall: 32.5 -> 30.6 | Hurt |
Closest papers by meaning
- End-to-End Spatio-Temporal Action Localisation with Video Transformers (CVPR 2024 · similarity 0.90)
- Spot On: Action Localization from Pointly-Supervised Proposals (ECCV 2016 · similarity 0.90)
- Action Tubelet Detector for Spatio-Temporal Action Localization (ICCV 2017 · similarity 0.90)
- TubeR: Tubelet Transformer for Video Action Detection (CVPR 2022 · similarity 0.89)
- Unsupervised Action Discovery and Localization in Videos (ICCV 2017 · similarity 0.89)
- Recurrent Tubelet Proposal and Recognition Networks for Action Detection (ECCV 2018 · 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.
