This repository contains the materials I used to teach Generative Models in Computer Vision from August 2022 to December 2024, both at Pontificia Universidad Católica de Chile and in diploma programs and workshops at other institutions.

I typically taught one or two sessions per semester, during which I developed two distinct types of classes based on the learning objectives I wanted to achieve with students:

1. Historical Overview of Generative Computer Vision Models: Covering GANs, VAEs, Diffusion, Latent Diffusion, text-to-image generation, etc. This class is oriented toward professionals seeking to enter the AI industry, understanding it as a tool within a technology stack. Rather than focusing on mathematical details like the Kullback-Leibler divergence, I emphasize the historical progression of this field and how these models can be utilized in ML system development. This content is available at Modelos Generativos en CV.pdf.
2. Deep Dive into Diffusion Models: We begin by examining fundamental building blocks of this area—DDPMs, U-NETs, DDIM, Classifier-Free Guidance—before introducing modern formulations and techniques like DreamBooth and ControlNet. This class targets those who want a comprehensive understanding of the mathematics behind diffusion models and how modern models evolved, providing an informed foundation and up-to-date vocabulary for those looking to dive deeper into this area. This content is available at Difusion I and Difusion II.

Additionally, I created a notebook to familiarize attendees with the Diffusers library, demonstrating how straightforward it is to interact with state-of-the-art diffusion models.