Research map / Video / Action recognition
Open-VCLIP: Transforming CLIP to an Open-vocabulary Video Model via Interpolated Weight Optimization
Zejia Weng, Xitong Yang, Ang Li, Zuxuan Wu, Yu-Gang Jiang
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 temporal attention across neighboring video frames (ablation) | CLIP adapted as VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy | Improved |
| Add Interpolated Weight Regularization (IWR) during training (ablation) | Open-VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy | Improved |
| Add stochastic weight averaging (SWA) (ablation) | Open-VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy | Improved |
| Use IWR and SWA together rather than either alone (ablation) | Open-VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy | Improved |
| Replace IWR with simple L2 weight regularization (ablation) | VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy | Hurt |
| Fine-tune CLIP conventionally instead of using Open-VCLIP optimization (ablation) | CLIP · Kinetics-600 | zero-shot accuracy: 68.1 -> 65.9 | Hurt |
| Fine-tune CLIP conventionally instead of using Open-VCLIP optimization (ablation) | CLIP ViT-L/14 · HMDB-51 | zero-shot accuracy: 55.0 -> 51.9 | Hurt |
| Adapt with ST-Adapter modules while keeping CLIP parameters frozen (ablation) | CLIP ViT-B/16 · Kinetics-600 | zero-shot accuracy: 68.1 -> 60.2$\pm$1.8 | Hurt |
| Use a larger ViT-L/14 backbone rather than ViT-B/16 (ablation) | Open-VCLIP · UCF-101, HMDB-51, Kinetics-600 | zero-shot accuracy: 80.5 -> 87.9; 55.0 -> 58.3; 76.2 -> 81.1 | Improved |
| Increase the CLIP interpolation coefficient from 0.5 to 0.7 (ablation) | Open-VCLIP · MSR-VTT | text-to-video recall@1: 31.3 -> 33.2 | Improved |
| Increase the CLIP interpolation coefficient from 0.5 to 0.7 (ablation) | Open-VCLIP · Kinetics-400 | accuracy: 78.9 -> 73.4 | Hurt |
| Increase the CLIP interpolation coefficient from 0.5 to 0.7 (ablation) | Open-VCLIP · MSR-VTT | video-to-text recall@1: 33.6 -> 34.4 | 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.
- ST-Adapter — ST-Adapter: Parameter-Efficient Image-to-Video Transfer Learning (NeurIPS 2022) · better on 2 of 2 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
- Interpreting and Analysing CLIP's Zero-Shot Image Classification via Mutual Knowledge (NeurIPS 2024 · similarity 0.88)
- FROSTER: Frozen CLIP is A Strong Teacher for Open-Vocabulary Action Recognition (ICLR 2024 · similarity 0.88)
- A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation (ICLR 2024 · similarity 0.88)
- ViLT-CLIP: Video and Language Tuning CLIP with Multimodal Prompt Learning and Scenario-Guided Optimization (AAAI 2024 · similarity 0.88)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.87)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · 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.
