AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Revisiting Classifier: Transferring Vision-Language Models for Video Recognition

Wenhao Wu, Zhun Sun, Wanli Ouyang

AAAI 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Open-vocabulary alignment from VLMs · 100 citations (OpenAlex)

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 changedWhereMeasuredEffect
Replace a randomly initialized learnable head with a frozen CLIP text-embedding classifierCLIP ViT-B/16 video recognition · Kinetics-400, zero-shotaccuracy: 0.2 -> 54.2Improved
Replace a randomly initialized learnable head with a frozen CLIP text-embedding classifierCLIP ViT-B/16 video recognition · Kinetics-400, 2-shotaccuracy: 21.6 -> 65.3Improved
Replace a randomly initialized learnable head with a frozen CLIP text-embedding classifierCLIP ViT-B/16 video recognition · Kinetics-400, full-shotaccuracy: 75.3 -> 80.1Improved
Add semantic text embeddings as the frozen classifier rather than random normal weights (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 59.3 -> 81.5Improved
Add semantic text embeddings as the frozen classifier rather than orthogonal weights (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 59.4 -> 81.5Improved
Use CLIP text embeddings instead of LDA visual-statistic classifier weights (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 80.8 -> 81.5Improved
Use DistilBERT rather than CLIP to produce the text classifier (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 81.5 -> 81.4No clear effect
Replace TAP with temporal transformer modeling (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 80.1 -> 81.5Improved
Add temporal 1D convolution to the global class token (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 80.1 -> 80.4Improved
Use temporal 1D convolutions in ResNet stages (ablation)CLIP ResNet-50 video recognition · Kinetics-400Top-1 accuracy: 71.2 -> 67.2Hurt
Use multiple fixed templates rather than one fixed template (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 81.5 -> 80.9Hurt
Use a learnable prompt template rather than one fixed template (ablation)CLIP ViT-B/16 video recognition · Kinetics-400Top-1 accuracy: 81.5 -> 81.2No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy