Dynamic Sampling Networks for Efficient Action Recognition in Videos

By Yin-Dong Zheng, Zhaoyang Liu, Tong Lu, Limin Wang. 2020

85 citations recorded by the source corpus.

The existing action recognition methods are mainly based on clip-level classifiers such as two-stream CNNs or 3D CNNs, which are trained from the randomly selected clips and applied to densely sampled clips during testing. However, this standard setting might be suboptimal for training classifiers and also requires huge computational overhead when deployed in practice. To address these issues, we propose a new framework for action recognition in videos, called Dynamic Sampling Networks (DSN), by designing a dynamic sampling module to improve the discriminative power of learned clip-level classifiers and as well increase the inference efficiency during testing. Specifically, DSN is composed of a sampling module and a classification module, whose objective is to learn a sampling policy to on-the-fly select which clips to keep and train a clip-level classifier to perform action recognition based on these selected clips, respectively. In particular, given an input video, we train an observation network in an associative reinforcement learning setting to maximize the rewards of the selected clips with a correct prediction. We perform extensive experiments to study different aspects of the DSN framework on four action recognition datasets: UCF101, HMDB51, THUMOS14, and ActivityNet v1.3. The experimental results demonstrate that DSN is able to greatly improve the inference efficiency by only using less than half of the clips, which can still obtain a slightly better or comparable recognition accuracy to the state-of-the-art approaches.

Subfields: Artificial Intelligence, Biomedical Engineering, Computer Vision and Pattern Recognition, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Shiji Song
2. Larry S. Davis
3. Yulin Wang
4. Gao Huang
5. Zuxuan Wu
6. Hengduo Li
7. Abhinav Shrivastava
8. Le Yang

In Biomedical Engineering

1. Larry S. Davis
2. Abhinav Shrivastava
3. Hengduo Li
4. Amir Ghodrati

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Shiji Song
3. Larry S. Davis
4. Dacheng Tao
5. Abhinav Kumar Gupta
6. Xiaogang Wang
7. Gao Huang
8. Zuxuan Wu

In Efficiency & Compression

1. Shiji Song
2. Gao Huang
3. Yulin Wang
4. Hengduo Li
5. Zuxuan Wu
6. Le Yang
7. Yizeng Han
8. Honghui Wang

Citations in the corpus

• Dynamic Neural Networks: A Survey (2021, background): “Moreover, dynamic sampling network (DSN) [192] segments each video into multiple sections, and a sampling module with shared weights across the sections is exploited to sample one clip from each section.”
• 2D or not 2D? Adaptive 3D Convolution Selection for Efficient Video Recognition (2020, background): “This is an orthogonal yet complementary direction to existing work on fast video recognition, which either designs lightweight 3D architectures [35, 44, 8, 34] or develops clip selection schemes to use fewer clips for classification [43, 20, 12, 41, 48].”
• FrameExit: Conditional Early Exiting for Efficient Video Recognition (2021, background): “This is commonly done by training agents to find which frame to observe next [55, 52, 57, 8], ranking frames based on saliency [29], gating frames [25], skipping RNN states [4], or by using audio to select relevant frames [15, 29].”