Research map / Video / Action recognition
Transition Forests: Learning Discriminative Temporal Transitions for Action Recognition and Detection
Guillermo Garcia-Hernando, Tae-Kyun Kim
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 |
|---|---|---|---|
| Add transition-based training objective E_t alongside classification objective E_c (ablation) | Transition Forests · MSRC-12 | accuracy | Improved |
| Use transition probabilities together with static classification probabilities (ablation) | Transition Forests · MSRC-12 | accuracy | Improved |
| Initialize transition nodes with E_c rather than randomly (ablation) | Transition Forests · MSRC-12 | accuracy: +0.35% | Improved |
| Continue initialization comparison for twice as many iterations (ablation) | Transition Forests · MSRC-12 | accuracy: 0.07% | No clear effect |
| Increase ensemble size from 1 tree to 6 trees (ablation) | Transition Forests · MSRC-12 | accuracy: 86.42% -> 93.10% | Improved |
| Increase ensemble size from 6 trees to 12 trees (ablation) | Transition Forests · MSRC-12 | accuracy: 93.10% -> 94.22% | Improved |
| Replace JP frame representation with RJP (ablation) | Transition Forests · MSRC-12 | accuracy: 94.22 -> 97.54 | Improved |
| Replace JP frame representation with MP (ablation) | Transition Forests · MSRC-12 | accuracy: 94.22 -> 98.25 | Improved |
| Combine static pose cues and discriminative temporal transitions | Transition Forests · MSRC-12 | accuracy | Improved |
| Combine static pose cues and temporal information | Transition Forests · Online Action Detection | overall F1-score: 0.653 -> 0.712 | Improved |
| Use single-frame RF instead of temporally informed methods | Online Action Detection · Online Action Detection | overall F1-score: 0.578 | Mixed |
Closest papers by meaning
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.87)
- Adaptive RNN Tree for Large-Scale Human Action Recognition (ICCV 2017 · similarity 0.87)
- Ske2Grid: Skeleton-to-Grid Representation Learning for Action Recognition (ICML 2023 · similarity 0.86)
- UCF-STAR: A Large Scale Still Image Dataset for Understanding Human Actions (AAAI 2020 · similarity 0.86)
- Weakly-Supervised Action Transition Learning for Stochastic Human Motion Prediction (CVPR 2022 · similarity 0.86)
- Recovering Complete Actions for Cross-dataset Skeleton Action Recognition (NeurIPS 2024 · similarity 0.85)
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.
