Research map / Video / Action recognition
Language-based Action Concept Spaces Improve Video Self-Supervised Learning
Kanchana Ranasinghe, Michael S Ryoo
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 category-space concept distillation over CLIP with temporal modification (ablation) | LSS · HMDB-51 linear probing | top-1 acc: 67.3 -> 68.5 | Improved |
| add description-space concept distillation (ablation) | LSS · HMDB-51 linear probing | top-1 acc: 68.5 -> 69.0 | Improved |
| add concept alignment (ablation) | LSS · HMDB-51 linear probing | top-1 acc: 69.0 -> 69.2 | Improved |
| add concept alignment (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 49.2 -> 49.5 | Improved |
| add category-space concept distillation (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 47.2 -> 48.8 | Improved |
| generic SSL (SVT) from CLIP init instead of language-aligned SSL (ablation) | LSS · HMDB-51 linear probing | top-1 acc: 63.9 -> 62.2 | Hurt |
| remove uniform distribution prior (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 48.4 -> 33.4 | Hurt |
| remove uniform distribution prior (ablation) | LSS · UCF-101 zero-shot | top-1 acc: 71.1 -> 54.3 | Hurt |
| remove significance weight (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 48.4 -> 47.2 | Hurt |
| remove significance weight (ablation) | LSS · UCF-101 zero-shot | top-1 acc: 71.1 -> 70.3 | Hurt |
| add 2000 WordNet words to category space (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 49.5 -> 48.6 | Hurt |
| add 10K WordNet sentences to description space (ablation) | LSS · HMDB-51 zero-shot | top-1 acc: 49.5 -> 49.6 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoMAE — VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training (NeurIPS 2022) · better on 2 of 2 shared results
- VATT — VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Text (NeurIPS 2021) · better on 1 of 1 shared result
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
- Kronecker Mask and Interpretive Prompts are Language-Action Video Learners (ICLR 2025 · similarity 0.89)
- Align Before Adapt: Leveraging Entity-to-Region Alignments for Generalizable Video Action Recognition (CVPR 2024 · similarity 0.88)
- Efficient Transfer Learning for Video-language Foundation Models (CVPR 2025 · similarity 0.87)
- OST: Refining Text Knowledge with Optimal Spatio-Temporal Descriptor for General Video Recognition (CVPR 2024 · similarity 0.87)
- BDC-CLIP: Brownian Distance Covariance for Adapting CLIP to Action Recognition (ICML 2025 · similarity 0.87)
- SRTube: Video-Language Pre-Training with Action-Centric Video Tube Features and Semantic Role Labeling (CVPR 2024 · similarity 0.86)
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.
