Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Expanding Language-Image Pretrained Models for General Video Recognition

About

Contrastive language-image pretraining has shown great success in learning visual-textual joint representation from web-scale data, demonstrating remarkable "zero-shot" generalization ability for various image tasks. However, how to effectively expand such new language-image pretraining methods to video domains is still an open problem. In this work, we present a simple yet effective approach that adapts the pretrained language-image models to video recognition directly, instead of pretraining a new model from scratch. More concretely, to capture the long-range dependencies of frames along the temporal dimension, we propose a cross-frame attention mechanism that explicitly exchanges information across frames. Such module is lightweight and can be plugged into pretrained language-image models seamlessly. Moreover, we propose a video-specific prompting scheme, which leverages video content information for generating discriminative textual prompts. Extensive experiments demonstrate that our approach is effective and can be generalized to different video recognition scenarios. In particular, under fully-supervised settings, our approach achieves a top-1 accuracy of 87.1% on Kinectics-400, while using 12 times fewer FLOPs compared with Swin-L and ViViT-H. In zero-shot experiments, our approach surpasses the current state-of-the-art methods by +7.6% and +14.9% in terms of top-1 accuracy under two popular protocols. In few-shot scenarios, our approach outperforms previous best methods by +32.1% and +23.1% when the labeled data is extremely limited. Code and models are available at https://aka.ms/X-CLIP

Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling• 2022

Related benchmarks

TaskDatasetResultRank
Action RecognitionKinetics-400
Top-1 Acc84.7
413
Action RecognitionUCF101--
365
Action RecognitionSomething-Something v2 (test)
Top-1 Acc10
333
Action RecognitionUCF101 (test)
Accuracy96.3
307
Action RecognitionHMDB51 (test)
Accuracy0.717
249
Action RecognitionKinetics 400 (test)
Top-1 Accuracy84.7
245
Action RecognitionHMDB51
Top-1 Acc44.6
225
Video Action RecognitionKinetics-400
Top-1 Acc87.7
184
Video ClassificationSomething-Something v2 (test)
Top-1 Acc0.578
169
Video Action RecognitionUCF101
Top-1 Acc92
153
Showing 10 of 106 rows
...

Other info

Code

Follow for update