← Return to program

Learning to Sketch with Differentiable Rendering

Sunday 2:10 PM–2:40 PM in Hall A

Drawing (or rendering) has long been one of the surprising and amazing things computers can do. But what about sketching, are computers any good at it? Differentiable Rendering is a technique borrowed from the field of machine learning that does exactly that! In this talk we will cover how it works and demo what it can do.

See this talk and many more by getting your ticket to PyCon AU now!

I want a ticket!

Differentiable Rendering is a technique that takes the standard machine learning pipeline and utilizes it to generate images from vector information such as meshes, curves and polygons. The twist is that the "Differentiable" in the name means standard machine learning algorithms such as gradient descent can then be used to learn how a given image can be created from these building blocks! We will cover how to use PyTorch and Python to implement such a renderer with the help of automatic gradients, the principles behind how it works and a few things that it can do, besides covering what could be built on top of it and how it could help artists.

Felipe Tavares He/him • @felipe@social.treehouse.systems

I love all things computing, from satellite communication systems to machine learning. I learned about Python around 2011, when trying to embed it in a Java game engine, and after having worked professionally in machine learning where Python is the default language, it never left my toolbox. Besides arguing with computers, I spend my time making music and digital art.