An Efficient 3D CNN for Action/Object Segmentation in Video

TL;DR

This paper introduces an efficient end-to-end 3D CNN that combines spatial and temporal features for video object and action segmentation, significantly reducing computational costs while outperforming existing methods.

Contribution

It presents a novel 3D separable convolution approach within an encoder-decoder framework for simultaneous spatial-temporal segmentation in videos.

Findings

Outperforms state-of-the-art methods on multiple datasets

Reduces computational complexity with 3D separable convolutions

Effective for both object and action segmentation

Abstract

Convolutional Neural Network (CNN) based image segmentation has made great progress in recent years. However, video object segmentation remains a challenging task due to its high computational complexity. Most of the previous methods employ a two-stream CNN framework to handle spatial and motion features separately. In this paper, we propose an end-to-end encoder-decoder style 3D CNN to aggregate spatial and temporal information simultaneously for video object segmentation. To efficiently process video, we propose 3D separable convolution for the pyramid pooling module and decoder, which dramatically reduces the number of operations while maintaining the performance. Moreover, we also extend our framework to video action segmentation by adding an extra classifier to predict the action label for actors in videos. Extensive experiments on several video datasets demonstrate the superior performance of the proposed approach for action and object segmentation compared to the state-of-the-art.