Sim2real transfer learning for 3D human pose estimation: motion to the rescue
This addresses the scalability issue in 3D human pose estimation by reducing reliance on hard-to-obtain real 3D labels, though it is incremental as it builds on existing neural-network approaches with a new pre-processing step.
The paper tackles the sim2real transfer problem for 3D human pose estimation by showing that using motion cues like optical flow and 2D keypoint motion enables models trained on synthetic data to perform on par with state-of-the-art methods trained on real data, achieving full 3D mesh recovery on the 3D Poses in the Wild benchmark.
Synthetic visual data can provide practically infinite diversity and rich labels, while avoiding ethical issues with privacy and bias. However, for many tasks, current models trained on synthetic data generalize poorly to real data. The task of 3D human pose estimation is a particularly interesting example of this sim2real problem, because learning-based approaches perform reasonably well given real training data, yet labeled 3D poses are extremely difficult to obtain in the wild, limiting scalability. In this paper, we show that standard neural-network approaches, which perform poorly when trained on synthetic RGB images, can perform well when the data is pre-processed to extract cues about the person's motion, notably as optical flow and the motion of 2D keypoints. Therefore, our results suggest that motion can be a simple way to bridge a sim2real gap when video is available. We evaluate on the 3D Poses in the Wild dataset, the most challenging modern benchmark for 3D pose estimation, where we show full 3D mesh recovery that is on par with state-of-the-art methods trained on real 3D sequences, despite training only on synthetic humans from the SURREAL dataset.