WebSight demonstrates a comparison of performance between JavaScript, asm.js, and WebAssembly. A user uploaded static image or live video is displayed for each target. Performance is measured by the length of time it takes to detect face(s) or eyes in the image or video.
Each target is run in its own web worker. The popular open source computer vision library, OpenCV, was compiled using Emscripten to asm.js and wasm (WebAssembly module) to utilize the Viola-Jones algorithm for object detection. HAAR.js was modified to remove the DOM manipulation, so it could be used in the JavaScript web worker.
Thanks to WebDSP for the inspiration, and to uscisysarch, njor, and foo123 for giving us a starting point.
Contributors to this project are BrianJFeldman, DebraD, MarkGeeRomano and YervantB.