MIT EECS - Amazon Undergraduate Research and Innovation Scholar
An Image Processing Application for iOS using Halide
With the proliferation of visual data, image processing pipelines have become essential in the handling of information from cameras and sensors. These pipelines must maintain high performance in the face of rapidly increasing image resolutions and complexity of image processing algorithms; but optimization of these pipelines can be painstaking, resulting in bulky and inflexible code. Halide is an image processing language which addresses these issues by decoupling the algorithm (what is being computed) from the schedule (where and when it’s computed). My work aims to create a key demo of this new technology, an iOS image editing application. The end result will push the envelope for Halide in terms of program size and help inform the future development of the technology.
I worked this past summer at Google training classifiers for Youtube videos and implementing a pipeline for classifier training. Previously, I’ve worked on building web applications using Ruby on Rails and other web technologies.