Intermediate Generative AI for Robotics Course - Python
Cutting edge Generative AI models applied to robotics.
Course Summary
This course offers an engaging dive into the world of Generative AI at an intermediate level, seamlessly blending theory with practical application. Students will explore advanced AI techniques such as imitation learning, diffusion-based navigation, and state-of-the-art transformer architectures, gaining both the knowledge and hands-on experience needed to tackle real-world challenges in robotics and beyond.
What you will learn
- Data Generation and Collection for Generative AI: Gain expertise in crafting and managing high-quality datasets to power advanced AI systems.
- Imitation Learning: Understand how AI models can replicate expert behavior to solve complex tasks across diverse applications.
- Diffusion-Based Models: Explore how diffusion techniques enable AI to predict outcomes, generate solutions, and adapt to dynamic challenges.
- Detection Transformers (DETR): Discover how transformers are revolutionizing object detection, unlocking new possibilities in computer vision.
- Vision Transformers (ViT): Learn how transformers analyze visual data to make intelligent, real-time decisions in sophisticated scenarios.
Course Overview
Demo
Demo unit where you will experience what can be achieved with Transformers applied to robotics
Imitation Learning
In this unit, you'll explore Imitation Learning, focusing on the Behavioral Cloning algorithm. You'll train a rover to navigate Mars using camera data and expert demonstrations, covering data collection, preparation, model training and deploying.
Diffusion Models
This unit covers diffusion models, a class of generative AI models, focusing on feature extraction, cosine similarity, and ROS 2 integration. Students will implement autonomous navigation using real-time image processing followed by model inference.
Transformers
This unit provides a comprehensive explanation of transformers used in generative AI applications. Special emphasis has been made on positional encodings
Real-Time Decision Making
This unit shows how vision transformers can be used to make a robot navigate in dynamic, unexplored environments.
CapStone Project
Capstone project where you will apply everything you have learned throughout the course.
Teachers
Arushi Khokhar
Generative AI
