Deep Learning has revolutionized AI, but it has also proven an extremely data hungry technology. In practicing Deep Learning, many have found that more data is better. Some see this as an insurmountable challenge as manually labelled training data becomes prohibitively costly and time-consuming. However, we see this as an opportunity to leverage real-time 3D video game technology to generate practically infinite amounts of synthetic training data whether for supervised learning in Computer Vision or unsupervised Reinforcement Learning. In this presentation we will explain how interactive large-scale synthetic data generation accelerates the AI revolution.

Download the slides for this talk.Download ( PDF, 57708.97 MB)