Learning Spatio-Temporal Features with 3D Residual Networks for Action Recognition
This work addresses action recognition for video analysis, representing an incremental improvement by applying known deep learning architectures to 3D CNNs.
The paper tackled the problem of action recognition in videos by proposing 3D ResNets, which adapt residual networks to 3D CNNs to extract spatio-temporal features, achieving better performance than shallow networks like C3D on datasets such as Kinetics and ActivityNet without overfitting despite large parameters.
Convolutional neural networks with spatio-temporal 3D kernels (3D CNNs) have an ability to directly extract spatio-temporal features from videos for action recognition. Although the 3D kernels tend to overfit because of a large number of their parameters, the 3D CNNs are greatly improved by using recent huge video databases. However, the architecture of 3D CNNs is relatively shallow against to the success of very deep neural networks in 2D-based CNNs, such as residual networks (ResNets). In this paper, we propose a 3D CNNs based on ResNets toward a better action representation. We describe the training procedure of our 3D ResNets in details. We experimentally evaluate the 3D ResNets on the ActivityNet and Kinetics datasets. The 3D ResNets trained on the Kinetics did not suffer from overfitting despite the large number of parameters of the model, and achieved better performance than relatively shallow networks, such as C3D. Our code and pretrained models (e.g. Kinetics and ActivityNet) are publicly available at https://github.com/kenshohara/3D-ResNets.