Demos Created using TensorFlow.JS

Facemesh-Camera.html :

        Real-time Face Tracking using TensorFlow.js

        Takes live video stream from webcam and creates facemesh
        (facial feature-map) by capturing 468 facial landmark points

        To run using .mp4 video file, simply add
        <source src="https://github.com/pamruta/TensorFlowJS/raw/master/VIDEO_PATH" type="video/mp4">
        tag inside <video> </video> in HTML

        and comment out live_webcam() function call..

        Link to TensorFlow.JS FaceMesh model:
        https://github.com/tensorflow/tfjs-models/tree/master/facemesh

Body-Segment.html :

        Body Segmentation using TensorFlow.JS Body-Pix model

        Takes live video stream from web camera, and runs
        segmentation using Body-Pix model from TensorFlow.JS

        Link to TensorFlow.JS Body-Pix model :

        https://github.com/tensorflow/tfjs-models/tree/master/body-pix

Video-Tagging.html :

        Identifies animal species in Wild-Life documentary films using
        Mobile-Net Image Classification model in TensorFlow.JS

        Link to TensorFlow.JS Mobile-Net model :

        https://github.com/tensorflow/tfjs-models/tree/master/mobilenet

        Sample video : wild-life.mp4

Object-Detection.html :

        Marks Bounding Boxes around the Objects detected in the video using
        TensorFlow.JS COCO-SSD model..

        Link to TensorFlow.JS COCO-SSD model :

        https://github.com/tensorflow/tfjs-models/tree/master/coco-ssd

Activity-Recognition.html :

        Builds a KNN Classifier for Sports Activity Recognition using
        Transfer Learning on a pre-trained model Mobile-Net

        Link to KNN Classifier in TensorFlow.JS :

        https://github.com/tensorflow/tfjs-models/tree/master/knn-classifier

        Activity Recognition can be performed in 3 simple steps:

        1. Compile training video which demonstrates various sports activities
        that you want to capture automatically using Machine Learning

        2. For each activity, capture few screenshots from the training video
        and tag them manually to train the knn-classifier

        3. Run predictions using the knn-classifier

Chroma-Key.html :

        Extracts person from an input video stream using Body-Pix model in
        TensorFlow.JS and replaces the video backdrop

        Sample background images are provided in 'backdrops' directory 

        Link to TensorFlow.JS Body-Pix model:

        https://github.com/tensorflow/tfjs-models/tree/master/body-pix

Hand-Tracking.html :

        Using Hand-Gestures to move objects in Virtual / Augmented Worlds

        Hand Positions are tracked using Hand-Pose model in TensorFlow.JS

        Link to Hand-Pose model:

        https://github.com/tensorflow/tfjs-models/tree/master/handpose

        More gesture-controlled features to Pause, Browse, Scroll, Zoom Media Stream to be added..

Sound-Box.html :

        Music Generation using Hand-Gestures

        Generates sounds using Tone.JS Java Script library, 
        based on Hand Positions captured by TensorFlow.JS Hand-Pose model..

        Each sound-box at (x, y) produces a sound with frequency f(x,y)

        Link to Tone.JS library:

        https://tonejs.github.io

Tech Stack: TensorFlow.JS, JavaScript, HTML5, Mobile-Net, Knn-Classifier, Coco-SSD, Tone.JS