AI Object Vision live demo

See what the model sees

Small AI models run directly in your browser via ONNX Runtime Web. Nothing is uploaded — camera, photos, and results stay on this device.

Choose a mode

Load only what you need. You can add another mode later.

Where should the model come from?

How it works

Classification

IMAGE → MODEL → LABEL + CONFIDENCE

Trained on millions of labeled images. It learns visual patterns that transfer to new photos — including yours.

Detection

IMAGE → MODEL → BOXES + LABELS

Finds regions, scores them against known classes, keeps the confident ones.

Stylize

IMAGE → MODEL → NEW IMAGE

Generative: the model outputs pixels, not a label. Same content, different artistic style.

All three run in this tab. Nothing is sent anywhere.
Model info
Mode—
Type—
InferenceBrowser (ONNX Runtime Web, WASM)
Loaded from—
Classes—
Load from this computer

Choose a single .onnx file. Nothing is uploaded.
Load from a URL

Public URL of a single .onnx file (must allow CORS).
Privacy