Unsupervised Learning of Object Structure and Dynamics from Videos
This work addresses the problem of extracting object-level motion understanding from videos without supervision, which is incremental as it builds on keypoint-based methods for video prediction and downstream tasks.
The paper tackles the challenge of unsupervised learning of object structure and dynamics from videos by using a keypoint-based representation and stochastic dynamics model, achieving improved performance over unstructured representations on tasks like object tracking and action recognition across diverse datasets.
Extracting and predicting object structure and dynamics from videos without supervision is a major challenge in machine learning. To address this challenge, we adopt a keypoint-based image representation and learn a stochastic dynamics model of the keypoints. Future frames are reconstructed from the keypoints and a reference frame. By modeling dynamics in the keypoint coordinate space, we achieve stable learning and avoid compounding of errors in pixel space. Our method improves upon unstructured representations both for pixel-level video prediction and for downstream tasks requiring object-level understanding of motion dynamics. We evaluate our model on diverse datasets: a multi-agent sports dataset, the Human3.6M dataset, and datasets based on continuous control tasks from the DeepMind Control Suite. The spatially structured representation outperforms unstructured representations on a range of motion-related tasks such as object tracking, action recognition and reward prediction.