Temporal Representation Learning on Monocular Videos for 3D Human Pose Estimation
This work addresses the problem of improving 3D human pose estimation from monocular videos for researchers and practitioners in computer vision, offering significant performance gains over existing unsupervised and weakly supervised methods.
This paper proposes an unsupervised feature extraction method for monocular videos to capture temporal information for 3D human pose estimation. It disentangles latent vectors into time-variant and time-invariant components, applying contrastive loss only to the time-variant features. This approach reduces error by approximately 50% compared to standard contrastive self-supervised strategies, outperforms other unsupervised single-view methods, and matches multi-view techniques.
In this paper we propose an unsupervised feature extraction method to capture temporal information on monocular videos, where we detect and encode subject of interest in each frame and leverage contrastive self-supervised (CSS) learning to extract rich latent vectors. Instead of simply treating the latent features of nearby frames as positive pairs and those of temporally-distant ones as negative pairs as in other CSS approaches, we explicitly disentangle each latent vector into a time-variant component and a time-invariant one. We then show that applying contrastive loss only to the time-variant features and encouraging a gradual transition on them between nearby and away frames while also reconstructing the input, extract rich temporal features, well-suited for human pose estimation. Our approach reduces error by about 50% compared to the standard CSS strategies, outperforms other unsupervised single-view methods and matches the performance of multi-view techniques. When 2D pose is available, our approach can extract even richer latent features and improve the 3D pose estimation accuracy, outperforming other state-of-the-art weakly supervised methods.