Skip-Clip: Self-Supervised Spatiotemporal Representation Learning by Future Clip Order Ranking
This addresses the annotation bottleneck for video analysis tasks, offering an incremental improvement over existing self-supervised methods.
The paper tackles the problem of reducing annotation requirements for training deep neural networks on videos by proposing Skip-Clip, a self-supervised method that learns spatiotemporal features through future clip order ranking, resulting in a 51.8% improvement over random initialization for action recognition on UCF101 and competitive performance with state-of-the-art self-supervision methods.
Deep neural networks require collecting and annotating large amounts of data to train successfully. In order to alleviate the annotation bottleneck, we propose a novel self-supervised representation learning approach for spatiotemporal features extracted from videos. We introduce Skip-Clip, a method that utilizes temporal coherence in videos, by training a deep model for future clip order ranking conditioned on a context clip as a surrogate objective for video future prediction. We show that features learned using our method are generalizable and transfer strongly to downstream tasks. For action recognition on the UCF101 dataset, we obtain 51.8% improvement over random initialization and outperform models initialized using inflated ImageNet parameters. Skip-Clip also achieves results competitive with state-of-the-art self-supervision methods.