Unsupervised Feature Learning from Temporal Data

Current state-of-the art object detection and recognition algorithms mainly use supervised training, and most benchmark datasets contain only static images. In this work we study feature learning in the context of temporally coherent video data. We focus on training convolutional features on unlabeled video data, using only the assumption that adjacent video frames contain semantically similar information. This assumption is exploited to train a pooling auto-encoder model regularized by slowness and sparsity. First, we confirm that fully connected networks mainly learn features stable under translation. Insipred by this observation, we proceed to train convolutional slow features which reveal richer invariants that are learned from natural video data.