Vision-Based Quadrotor Navigation

This lecture covers the final project presentation on quadrotor point navigation using deep reinforcement learning and low-level control. The project aims to train a quadrotor to navigate towards a target point in space, focusing on visual intelligence and the potential applications of quadrotors in various fields. The presentation discusses the problem definition, related work, the proposed approach, results, and conclusions. The project involves creating realistic face augmentations to improve gaze model robustness, with a focus on gaze estimation and its applications in social behavior analysis, human-computer interaction, and automotive safety.