MLVideo is a library that interactively and meaningfully visualizes outputs of vision models, such as pose detection.
MLVideo.mp4
A npm local package is provided under ./package.
To start a development server run npm install followed by npm run dev. The server will start on port 4000. The static files must be under ./playground.
To generate a new version of the local package run: npm run build-dist. Then, npm pack
A sample of the data is provided here: TBD
const mainDiv: HTMLDivElement = <HTMLDivElement>document.getElementById('div-container');
const video: MLVideo = new MLVideo( mainDiv, './path/to/video.mp4');
// poses variable has a set of poses per frame; Each pose is defined by a set of joints ([number, number]). We assume there will be 17 joints per skeleton, following the YOLO pose format.
const range: [number, number] = [5000, 6000];
let poses: number[][][] = await fetch('./path/to/people.json');
poses = poses.slice( range[0], range[1] );
video.set_range_data( range, poses );
video.play_range();