On the Benefits of 3D Pose and Tracking for Human Action Recognition
This work improves action recognition accuracy for video analysis applications, though it is incremental as it builds on existing tracking and fusion methods.
The paper tackles human action recognition by using 3D pose and tracking over trajectories, achieving state-of-the-art performance on the AVA v2.2 dataset with a +10.0 mAP gain for pose-only models and +2.8 mAP gain for fused models.
In this work we study the benefits of using tracking and 3D poses for action recognition. To achieve this, we take the Lagrangian view on analysing actions over a trajectory of human motion rather than at a fixed point in space. Taking this stand allows us to use the tracklets of people to predict their actions. In this spirit, first we show the benefits of using 3D pose to infer actions, and study person-person interactions. Subsequently, we propose a Lagrangian Action Recognition model by fusing 3D pose and contextualized appearance over tracklets. To this end, our method achieves state-of-the-art performance on the AVA v2.2 dataset on both pose only settings and on standard benchmark settings. When reasoning about the action using only pose cues, our pose model achieves +10.0 mAP gain over the corresponding state-of-the-art while our fused model has a gain of +2.8 mAP over the best state-of-the-art model. Code and results are available at: https://brjathu.github.io/LART