Deep Reinforcement Learning for Object Tracking

Explore deep reinforcement learning techniques for object tracking in this 30-minute lecture from the University of Central Florida. Learn about different types of learning, deep learning fundamentals, and reinforcement learning principles. Discover the concept of value functions as predictions of future rewards and various approaches to reinforcement learning. Delve into action-driven object tracking, including problem definition in an RL setting, action-decision networks, and training methods such as supervised learning, reinforcement learning, and online adaptation. Gain insights into self-comparison techniques and their applications in object tracking systems.

Syllabus

Intro
Different types of learning
Deep learning in a nutshell
Reinforcement learning in a nutshell
Policy
Value function A value function is a prediction of future reward
Approaches To Reinforcement Learning
Motivation
Action-driven object tracking
Problem definition (RL setting)
Action-decision network
Training: Supervised learning
Training: Reinforcement learning
Training: Online adaptation
Self comparison