In-Browser Inference Demos
A growing portfolio of machine-learning models running entirely in your browser — no server, no upload. WebGPU when available, optimized WASM everywhere else.
Computer Vision
The models, shown straight — detections and keypoints drawn over your input.
-
Object Detection
Detect and label objects in images or a live camera feed.
-
Pose Estimation
Detect people and their body keypoints from an image, video, or camera.
-
Action Recognition
Classify the human action in each frame — single-frame, smoothed over time on video.
-
Face Mesh & Orientation
A 468-point face mesh, head orientation, and live expressions (MediaPipe).
-
Eye Tracking
Track irises, eye-openness/blinks, and gaze direction from the webcam (MediaPipe).
-
Hand Tracking
Track 21 landmarks per hand, plus handedness, from an image or camera (MediaPipe).
-
Face Recognition
Enroll a face, then verify matches live by descriptor distance (face-api.js / TF.js).