Use Azure Custom Vision to Repurpose the Lunar Rover

Overview

Learn how to build an image classifier using Azure Custom Vision to identify objects in images, inspired by the Lunar Rover from the Netflix Original "Over the Moon". Follow along as Dr. G demonstrates how to create and train an AI model to detect a specific character (Bungee) in images. Explore the process of setting up an Azure Custom Vision resource, uploading and tagging images, training the AI model, and testing its performance. Discover best practices for image selection, the importance of tagging both positive and negative examples, and how to refine your model for better accuracy. No coding experience required, making this tutorial accessible for beginners interested in AI and image recognition technology.

Syllabus

- Introduction and Agenda.
- Goal: Use AI to detect objects in images.
- Create Azure Custom Vision Resource.
- Training vs Predicting with Custom Vision AI.
- How to find your resource in Azure.
- Quickstart guide for Custom Vision.
- Intro to Custom Vision Portal and Create Project.
- Upload image and tag it.
- Best images to use for Custom Vision.
- Upload and tag 14 more images, then train AI.
- Review results of training.
- Test AI with a new image.
- Test AI with a complex image.
- Why we should tag "not bungee" objects in the image.
- Tagging a images with "bungee" and "not bungee".
- Train and review AI model with the new images.
- Test AI with new images.
- Recap.