Expanding Language-Image Pretrained Models for General Video Recognition

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

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Zero-Shot Action Recognition HMDB51 X-CLIP Top-1 Accuracy 44.6 # 12
Zero-Shot Action Recognition Kinetics X-CLIP Top-1 Accuracy 65.2 # 9
Top-5 Accuracy 86.1 # 6
Action Classification Kinetics-400 X-CLIP(ViT-L/14, CLIP) Acc@1 87.7 # 27
Acc@5 97.4 # 24
Action Classification Kinetics-600 X-CLIP(ViT-L/14, CLIP) Top-1 Accuracy 88.3 # 20
Top-5 Accuracy 97.7 # 13
Zero-Shot Action Recognition UCF101 X-CLIP Top-1 Accuracy 72.0 # 13

Methods


No methods listed for this paper. Add relevant methods here