Self-Supervised Video Representation Learning with Space-Time Cubic Puzzles
arXiv:1811.09795
Abstract
Self-supervised tasks such as colorization, inpainting and zigsaw puzzle have been utilized for visual representation learning for still images, when the number of labeled images is limited or absent at all. Recently, this worthwhile stream of study extends to video domain where the cost of human labeling is even more expensive. However, the most of existing methods are still based on 2D CNN architectures that can not directly capture spatio-temporal information for video applications. In this paper, we introduce a new self-supervised task called as \textit{Space-Time Cubic Puzzles} to train 3D CNNs using large scale video dataset. This task requires a network to arrange permuted 3D spatio-temporal crops. By completing \textit{Space-Time Cubic Puzzles}, the network learns both spatial appearance and temporal relation of video frames, which is our final goal. In experiments, we demonstrate that our learned 3D representation is well transferred to action recognition tasks, and outperforms state-of-the-art 2D CNN-based competitors on UCF101 and HMDB51 datasets.
arXiv commentAccepted to AAAI 2019
References in corpus (10)
- Two-Stream Convolutional Networks for Action Recognition in Videos
- UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild
- The Kinetics Human Action Video Dataset
- Generating Videos with Scene Dynamics
- Temporal Segment Networks: Towards Good Practices for Deep Action Recognition
- Unsupervised Visual Representation Learning by Context Prediction
- Unsupervised Learning of Visual Representations using Videos
- Improvements to context based self-supervised learning
- Two-Phase Learning for Weakly Supervised Object Localization
- Learning Image Representations by Completing Damaged Jigsaw Puzzles