Learning Multi-Human Optical Flow
This work addresses the lack of training data for human optical flow, benefiting computer vision researchers and applications in human action analysis.
The authors tackled the problem of estimating optical flow for human motion by creating a synthetic dataset of multi-human optical flow and training deep networks on it, resulting in networks that outperformed top methods on test data and generalized well to real sequences.
The optical flow of humans is well known to be useful for the analysis of human action. Recent optical flow methods focus on training deep networks to approach the problem. However, the training data used by them does not cover the domain of human motion. Therefore, we develop a dataset of multi-human optical flow and train optical flow networks on this dataset. We use a 3D model of the human body and motion capture data to synthesize realistic flow fields in both single- and multi-person images. We then train optical flow networks to estimate human flow fields from pairs of images. We demonstrate that our trained networks are more accurate than a wide range of top methods on held-out test data and that they can generalize well to real image sequences. The code, trained models and the dataset are available for research.