ViViT: A Video Vision Transformer

We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatio-temporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks. To facilitate further research, we release code at

PDF Abstract ICCV 2021 PDF ICCV 2021 Abstract

Results from the Paper

Ranked #8 on Action Classification on MiT (Top 5 Accuracy metric, using extra training data)

     Get a GitHub badge
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Action Recognition EPIC-KITCHENS-100 ViViT-L/16x2 Fact. encoder Action@1 44.0 # 20
Verb@1 66.4 # 21
Noun@1 56.8 # 17
Action Classification Kinetics-400 ViViT-L/16x2 (JFT) Acc@1 84.9 # 53
Acc@5 95.8 # 43
Action Classification Kinetics-400 ViViT-L/16x2 320 Acc@5 94.7 # 59
Action Classification Kinetics-600 ViViT-H/16x2 (JFT) Top-1 Accuracy 85.8 # 29
Top-5 Accuracy 96.5 # 23
Action Classification Kinetics-600 ViViT-L/16x2 Top-1 Accuracy 84.3 # 33
Top-5 Accuracy 95.6 # 37
Action Classification MiT ViViT-L/16x2 Top 5 Accuracy 64.9 # 8
Action Recognition Something-Something V2 ViViT-L/16x2 Fact. encoder Top-1 Accuracy 65.4 # 87
Top-5 Accuracy 89.8 # 63

Results from Other Papers

Task Dataset Model Metric Name Metric Value Rank Uses Extra
Training Data
Source Paper Compare
Action Classification Kinetics-600 ViViT-L/16x2 (320x320) Top-1 Accuracy 83.0 # 42
Top-5 Accuracy 95.7 # 33


No methods listed for this paper. Add relevant methods here