Research map / Video / Action recognition
XOV-Action: Toward Generalizable Open-Vocabulary Action Recognition
Kun-Yu Lin, Henghui Ding, Jia-Run Du, Jiaming Zhou, Yi-Xing Peng, Yu-Ming Tang, Zhilin Zhao, Chen Change Loy, Wei-Shi Zheng
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 |
|---|---|---|---|
| add scene-aware discrimination loss L_scene (ablation) | L_vta baseline · XOVABench (trained on Kinetics150) ARID/NEC-Dr | closed-set accuracy: +1.83% and +4.63% | Improved |
| add L_scene alone (ablation) | L_vta baseline · XOVABench (trained on Kinetics150) average | open-set accuracy: 38.64 -> 37.00 | Hurt |
| add L_scene alone (ablation) | L_vta baseline · XOVABench (trained on Kinetics150) average | overall accuracy: 46.94 -> 47.42 | Improved |
| add action-aware discrimination loss L_action (ablation) | L_vta + L_scene · XOVABench (trained on Kinetics150) average | open-set accuracy: 37.00 -> 39.21 | Improved |
| add diversified elaboration representation learning (ablation) | L_vta + L_scene + L_action · XOVABench (trained on Kinetics150) average | overall accuracy: 48.00 -> 49.80 | Improved |
| add diversified elaboration representation learning (ablation) | L_vta + L_scene + L_action · XOVABench (trained on Kinetics150) average | open-set accuracy: 39.21 -> 41.79 | Improved |
| use textual descriptions at inference only (ablation) | XOV-Action without elaboration learning · XOVABench (trained on Kinetics150) average | open-set accuracy: 39.21 -> 40.56 | Improved |
| add video elaboration modules (ablation) | XOV-Action · XOVABench (trained on Kinetics150) average | open-set accuracy: 40.85 -> 41.79 | Improved |
| add adaptive elaboration matching (ablation) | XOV-Action · XOVABench (trained on Kinetics150) average | open-set accuracy: 41.13 -> 41.79 | Improved |
| add adaptive elaboration matching (ablation) | XOV-Action · XOVABench (trained on Kinetics150) average | closed-set accuracy: 54.97 -> 55.96 | Improved |
| add L_scene (ablation) | XOV-Action · XOVABench (trained on Kinetics150) average | closed-set accuracy: 53.85 -> 55.96 | Improved |
| add L_action (ablation) | XOV-Action · XOVABench (trained on Kinetics150) average | open-set accuracy: 40.99 -> 41.79 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FROSTER — FROSTER: Frozen CLIP is A Strong Teacher for Open-Vocabulary Action Recognition (ICLR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FROSTER: Frozen CLIP is A Strong Teacher for Open-Vocabulary Action Recognition (ICLR 2024 · similarity 0.90)
- Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection? (NeurIPS 2024 · similarity 0.89)
- Open-Vocabulary Video Relation Extraction (AAAI 2024 · similarity 0.89)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (NeurIPS 2025 · similarity 0.89)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.88)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.87)
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.
