Writing Fast Image-Processing Code With Halide


Writing Fast Image-Processing Code With Halide

Tuesday, 11 August 10:45 AM - 12:15 PM | Los Angeles Convention Center, Room 404AB

Halide is a programming language for image processing and computational photography that is now in widespread use at companies and research labs. It has been used to create state-of-the-art code including Google's HDR+ pipeline on Glass and millions of Android phones, and Google Photos AutoEnhance feature, which runs on tens of thousands of data-center machines.

This course introduces programmers from the graphics community to the core concepts in Halide and explains how to use Halide to productively write high-performance image-processing, computational-photography, and vision code. The course introduces the major concepts in Halide, shows how to set up a Halide environment, summarizes writing and compiling basic programs, and demonstrates how to implement and optimize a non-trivial image-processing pipeline, all in a short course and very few lines of code.

Course Schedule

10:45 am
Introduction to Halide

11:05 am
Writing Your First Halide program

11:20 am
Optimizing Gaussian blur from scratch
Adams & Ragan-Kelley

11:45 am
Optimizing an image enhancement pipeline
Adams & Ragan-Kelley




Programming experience.

Intended Audience

Graphics and vision programmers and researchers interested in building high-performance image processing pipelines using Halide.


Jonathan Ragan-Kelley
Stanford University

Andrew Adams
Google Inc.

Dillon Sharlet
Google Inc.

Frédo Durand
Massachusetts Institute of Technology