Saliency Tubes: Visual Explanations for Spatio-Temporal Convolutions
This addresses the problem of interpretability in complex 3D CNNs for video analysis, providing a visualization tool for researchers and practitioners, but it is incremental as it builds on existing saliency methods.
The authors tackled the challenge of visualizing informative spatio-temporal regions in 3D convolutional neural networks for video classification, proposing Saliency Tubes to highlight key points and regions over time, and demonstrated this on action classification datasets.
Deep learning approaches have been established as the main methodology for video classification and recognition. Recently, 3-dimensional convolutions have been used to achieve state-of-the-art performance in many challenging video datasets. Because of the high level of complexity of these methods, as the convolution operations are also extended to additional dimension in order to extract features from them as well, providing a visualization for the signals that the network interpret as informative, is a challenging task. An effective notion of understanding the network's inner-workings would be to isolate the spatio-temporal regions on the video that the network finds most informative. We propose a method called Saliency Tubes which demonstrate the foremost points and regions in both frame level and over time that are found to be the main focus points of the network. We demonstrate our findings on widely used datasets for third-person and egocentric action classification and enhance the set of methods and visualizations that improve 3D Convolutional Neural Networks (CNNs) intelligibility.