Affiliation:
1. Jiangnan University, Wuxi, China
2. Suzhou University of Science and Technology, Suzhou, China
3. University of Surrey, Guildford, UK
Abstract
Video action recognition aims at classifying the action category in given videos. In general, semantic-relevant video frame pairs reflect significant action patterns such as object appearance variation and abstract temporal concepts like speed, rhythm, and so on. However, existing action recognition approaches tend to holistically extract spatiotemporal features. Though effective, there is still a risk of neglecting the crucial action features occurring across frames with a long-term temporal span. Motivated by this, in this article, we propose to perceive actions via frame pairs directly and devise a novel Nest Structure with frame pairs as basic units. Specifically, we decompose a video sequence into all possible frame pairs and hierarchically organize them according to temporal frequency and order, thus transforming the original video sequence into a Nest Structure. Through naturally decomposing actions, the proposed structure can flexibly adapt to diverse action variations such as speed or rhythm changes. Next, we devise a Temporal Pair Analysis module (TPA) to extract discriminative action patterns based on the proposed Nest Structure. The designed TPA module consists of a pair calculation part to calculate the pair features and a pair fusion part to hierarchically fuse the pair features for recognizing actions. The proposed TPA can be flexibly integrated into existing backbones, serving as a side branch to capture various action patterns from multi-level features. Extensive experiments show that the proposed TPA module can achieve consistent improvements over several typical backbones, reaching or updating CNN-based SOTA results on several challenging action recognition benchmarks.
Funder
National Key Research and Development Program of China
National Natural Science Foundation of China
111 Project of Ministry of Education of China
Engineering and Physical Sciences Research Council
Publisher
Association for Computing Machinery (ACM)
Reference68 articles.
1. Humam Alwassel, Silvio Giancola, and Bernard Ghanem. 2021. Tsp: Temporally-sensitive pretraining of video encoders for localization tasks. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 3173–3183.
2. Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset
3. Junyoung Chung Caglar Gulcehre Kyunghyun Cho and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. In NIPS 2014 Workshop on Deep Learning.
4. Iulia Duta Andrei Nicolicioiu and Marius Leordeanu. 2020. Dynamic regions graph neural networks for spatio-temporal reasoning. In ORLR Workshop Neural Information Processing Systems (NeurIPS’20).
5. Nour Eldin Elmadany Yifeng He and Ling Guan. 2021. Improving action recognition via temporal and complementary learning. ACM Transactions on Intelligent Systems and Technology 12 3 (2021) 1–24.