Blog

Diffusion Methods for Robotics

CMSC818B Decision Making for Robotics Mini-Project 1 Pranav Asthana (UID: 121089670) 1 Overview Generative models have been popular for many tasks in vision and language for several years. Since about 2015, diffusion models have become popular for image generation tasks like inpainting, superresolution, denoising, stylistic editing, relighting, text-to-image synthesis, and several other conditional image generation tasks. Their expressivity and ability to combine different concepts made them very popular over the previous generative methods such as Generative Adversarial Networks (GANs). Diffusion...

Visualizing evolution in code

In my first year as an undergrad CS student, I was fascinated by a lot of things. Evolutionary or genetic algorithms was one such fascination that lasted long enough for me to write a system that evolves to reach a goal. In this post, I’m going to explain the idea behind applying Darwin’s theory of evolution to solve optimization problems and am going to show that we can visualize how they work. This post is for: People who are fascinated...