AI Paper Atlas

Research map / Video / Video-language understanding

Frozen CLIP Models Are Efficient Video Learners

Ziyi Lin, Shijie Geng, Renrui Zhang, Peng Gao, Gerard de Melo, Xiaogang Wang, Jifeng Dai, Yu Qiao, Hongsheng Li

ECCV 2022 · Video › Video-language understanding · approach: Video-language pretraining · 242 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
use 4 decoder blocks instead of 1 (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 81.1 -> 82.9Improved
feed decoder blocks multi-layer intermediate features [-4,-3,-2,-1] instead of last layer only (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 82.1 -> 82.9Improved
use spatiotemporal unpooled features instead of temporal-only token features (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 79.8 -> 82.9Improved
use spatiotemporal features instead of spatial-only averaged features (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 80.1 -> 82.9Improved
add temporal modules (T-Conv, T-PE, T-CA) (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 82.5 -> 82.9No clear effect
add temporal modules (T-Conv, T-PE, T-CA) (ablation)EVL ViT-B/16 · Something-Something-v2top-1 accuracy: 47.2 -> 61.0Improved
use 12 decoder blocks instead of 4 (ablation)EVL ViT-B/16 · Something-Something-v2top-1 accuracy: 58.6 -> 61.0Improved
increase decoder depth from 4 to 5 (ablation)EVL ViT-B/16 · Kinetics-400top-1 accuracy: 82.9 -> 83.0No clear effect
freeze CLIP backbone vs fully finetune (CLIP pretrain)EVL vs TimeSformer DividedST · Kinetics-400top-1 accuracy: 82.0 -> 82.9Improved
freeze backbone vs fully finetune (ImageNet-21k pretrain)EVL vs TimeSformer DividedST · Kinetics-400top-1 accuracy: 78.0 -> 75.4Hurt
ensemble EVL ViT-B/16 (8) with Uniformer-B (32)Uniformer-B · Kinetics-400top-1 accuracy: 82.9 -> 85.2Improved
ensemble EVL (32) with Uniformer-B (32)Uniformer-B · Something-Something-v2top-1 accuracy: 71.2 -> 72.1Improved

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