A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning
This work addresses the problem of learning effective video representations without labeled data, which is incremental as it builds on existing image-based frameworks but extends them to spatiotemporal domains.
The paper tackles unsupervised spatiotemporal representation learning from videos by introducing a simple objective that encourages temporally-persistent features, achieving state-of-the-art results in multiple benchmarks and showing cases where unsupervised pre-training outperforms supervised methods.
We present a large-scale study on unsupervised spatiotemporal representation learning from videos. With a unified perspective on four recent image-based frameworks, we study a simple objective that can easily generalize all these methods to space-time. Our objective encourages temporally-persistent features in the same video, and in spite of its simplicity, it works surprisingly well across: (i) different unsupervised frameworks, (ii) pre-training datasets, (iii) downstream datasets, and (iv) backbone architectures. We draw a series of intriguing observations from this study, e.g., we discover that encouraging long-spanned persistency can be effective even if the timespan is 60 seconds. In addition to state-of-the-art results in multiple benchmarks, we report a few promising cases in which unsupervised pre-training can outperform its supervised counterpart. Code is made available at https://github.com/facebookresearch/SlowFast