Computer Vision Research at NVIDIA: Using AI to Perceive Humans

Perceiving humans, such as their pose, gaze, and so on, is of great interest for many practical applications, including human-machine interaction, activity recognition, video analytics, visual effects, gaming, and any other application involving humans in the scenes. I'll present our work focusing on a detailed understanding of the human body from monocular RGB images. Specifically, I'll present deep learning methods for body pose estimation, hand pose estimation, fiducial landmark detection, and eye gaze estimation and redirection. We'll show how deep learning models can be made robust to face challenges posed by real-world scenarios and address the problem of data scarcity for training these methods.