Our new CVPR'19 Human Pose Estimation code is available on-line!
#OpenScience
We share the source code of our new method for multi-person 2D human pose estimation that is particularly well suited for urban mobility such as self-driving cars and delivery robots. Our method outperforms state-of-the-art work (e.g., from Facebook or Google) at low resolution and in crowded, cluttered and occluded scenes.
References
S. Kreiss, L. Bertoni, A. Alahi, "PifPaf: Composite Fields for Human Pose Estimation" in IEEE conference on computer vision and pattern recognition (CVPR'19).