Real computer-vision models, running live in your browser — no uploads, no servers. This is the same on-device inference we deploy to the edge for clients.
100% in-browser · your camera and images never leave your device
Point your camera at anything. The model finds and labels objects frame-by-frame — the same building block behind defect detection, inventory counting, and security analytics.
TensorFlow.js · COCO-SSD (80 object classes) · Runs entirely in your browser — your camera feed never leaves your device.
Click “Start camera” to run this model live on your device.
Separates people from the background pixel-by-pixel and blurs everything else — the basis for privacy masking, background replacement, and region-of-interest analytics.
TensorFlow.js · BodyPix segmentation · Runs entirely in your browser — your camera feed never leaves your device.
Click “Start camera” to run this model live on your device.
Detects faces and key facial landmarks in real time — the foundation for occupancy, attention, and people-flow analytics. This demo detects faces only; it does not identify or recognise anyone.
TensorFlow.js · BlazeFace (detection only, no identification) · Runs entirely in your browser — your camera feed never leaves your device.
Click “Start camera” to run this model live on your device.
Upload any photo and the model estimates how far away each pixel is — turning a flat image into a depth map. The basis for 3D reconstruction, robotics, and spatial analytics.
Transformers.js · Depth-Anything V2 (ONNX) · Runs entirely in your browser — your image never leaves your device.