UniFormerV2 in Keras

We have implemented UniFormerV2, a strong video networks achieved state-of-the-art recognition performance on 8 popular video benchmarks, including Kinetics-400/600/700, Moments in Time, Something-Something V1/V2, ActivityNet and HACS.

UniFormerV2 is the first model to achieve 90% top-1 accuracy on Kinetics-400. Total 35 (x2) checkpoints are now available in both SavedModel and H5 formats.

1 Like

Starter code to fine-tune on custom dataset.

UFV2-keras on Huggingfacae space

1 Like