Research map / Video / Action recognition
Unsupervised Action Discovery and Localization in Videos
Khurram Soomro, Mubarak Shah
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 |
|---|---|---|---|
| use dominant-set discriminative clustering instead of spectral clustering | action discovery · UCF Sports (iDTF features) | clustering accuracy: 59.4 -> 69.9 | Improved |
| use dominant-set discriminative clustering instead of spectral clustering | action discovery · UCF101 | clustering accuracy: 51.6 -> 61.2 | Improved |
| use dominant-set discriminative clustering instead of DAKM | action discovery · JHMDB | clustering accuracy: 50.2 -> 53.7 | Improved |
| use dominant-set discriminative clustering instead of DAKM | action discovery · UCF101 | clustering accuracy: 37.1 -> 61.2 | Improved |
| use dominant set alone (DS) instead of dominant sets plus iterative SVM selection | action discovery · UCF Sports (iDTF features) | clustering accuracy: 53.3 -> 69.9 | Improved |
| use knapsack annotation with perfect labels instead of Ma et al. weakly supervised method | action annotation · UCF Sports (average over actions) | localization AUC: 41.0% -> 47.7% | Improved |
| knapsack annotation vs Ma et al. on Ride | action annotation · UCF Sports (Ride) | localization AUC: 30.6% -> 14.1% | Hurt |
| knapsack annotation vs Ma et al. on Kick | action annotation · UCF Sports (Kick) | localization AUC: 48.3% -> 37.7% | Hurt |
| use C3D features instead of iDTF features for action discovery | action discovery · UCF Sports | clustering accuracy: 69.9 -> 90.1 | Improved |
| use C3D features instead of iDTF features for localization | unsupervised action localization · UCF Sports | localization AUC | No clear effect |
| add joint annotation selection (BIQP) across videos in a cluster (ablation) | unsupervised action localization · UCF Sports | localization AUC | Improved |
| add pairwise structural SVM term to knapsack action distinctness score (ablation) | knapsack localization · UCF Sports | localization AUC | Improved |
Closest papers by meaning
- Spot On: Action Localization from Pointly-Supervised Proposals (ECCV 2016 · similarity 0.91)
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · similarity 0.90)
- A flexible model for training action localization with varying levels of supervision (NeurIPS 2018 · similarity 0.90)
- ActionBytes: Learning From Trimmed Videos to Localize Actions (CVPR 2020 · similarity 0.90)
- Pointly-Supervised Action Localization (IJCV 2018 · similarity 0.90)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.88)
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.
