Spatiotemporal Residual Networks for Video Action Recognition

TL;DR

This paper introduces spatiotemporal Residual Networks that combine two-stream ConvNets and ResNets to improve video action recognition by enabling better spatiotemporal interaction and hierarchical feature learning.

Contribution

The paper presents a novel architecture that integrates residual connections into two-stream ConvNets for spatiotemporal modeling, enhancing action recognition performance.

Findings

Outperforms previous state-of-the-art on benchmark datasets

Effectively models spatiotemporal interactions in videos

Enables end-to-end training of deep spatiotemporal networks

Abstract

Two-stream Convolutional Networks (ConvNets) have shown strong performance for human action recognition in videos. Recently, Residual Networks (ResNets) have arisen as a new technique to train extremely deep architectures. In this paper, we introduce spatiotemporal ResNets as a combination of these two approaches. Our novel architecture generalizes ResNets for the spatiotemporal domain by introducing residual connections in two ways. First, we inject residual connections between the appearance and motion pathways of a two-stream architecture to allow spatiotemporal interaction between the two streams. Second, we transform pretrained image ConvNets into spatiotemporal networks by equipping these with learnable convolutional filters that are initialized as temporal residual connections and operate on adjacent feature maps in time. This approach slowly increases the spatiotemporal receptive field as the depth of the model increases and naturally integrates image ConvNet design principles. The whole model is trained end-to-end to allow hierarchical learning of complex spatiotemporal features. We evaluate our novel spatiotemporal ResNet using two widely used action recognition benchmarks where it exceeds the previous state-of-the-art.