Runs entirely in your browser

EV charging connector segmentation

A YOLO segmentation model identifies five connector types — AC Type 1 and 2, DC CCS Type 1 and 2, and CHAdeMO — from your camera, an image, or a video file. Inference runs on-device through onnxruntime-web: nothing is uploaded, and it keeps working offline once the model is cached.

Detection

/detection

Per-frame instance segmentation. Each frame is scored on its own, with no memory of the last one.

·Masks and boxes per frame
·Lowest latency
·Best for single images

Open

Detection + Tracking

/tracking

The same segmentation, plus identities that persist across frames — so a connector stays the same object as it moves.

·Stable IDs and a distinct-object count
·Class decided by a vote across frames, not one guess
·Kalman-smoothed boxes, tunable live

Open

Five models

YOLO26, YOLO12, YOLOv8, YOLO11m and YOLOv9c, switchable at run time. All export to one output layout, so the browser has a single decode path.

WebGPU when available

The runtime is picked per browser: WebGPU where it exists, WASM otherwise, and only the chosen build is downloaded.

Cached after first load

Model weights are stored in Cache Storage, so reloads and switching back to a model skip the download entirely.