This year at CVPR 2020 the team organized a tutorial on Disentangled Representations for Relightable Performance Capture of Humans. In this tutorial we will show how to build performance capture systems to learn representations that can be used for photorealistic renderings of humans in virtual and real scenes.
Photorealistic renderings are the key to solve a variety of perception tasks using Machine Learning. For instance, using The Relightables, we trained a deep convolutional network to enhance the depth sensor we built for Google Pixel 4. In this tutorial we will show many of such examples as well as how to push the state-of-art in performance capture.
Related to this, our team is also co-organizing the Neural Rendering tutorial with many friends and collaborators. This tutorial is based on our Eurographics work, and we will show many exciting applications in this new field.
I hope you enjoy these talks!