Vision Lab

This is the room where pixels tell their story. Test a browser vision model, teach a tiny classifier your own labels, and see which images cluster together in feature space.

Security reminderThis studio is for education and experimentation. Do not upload production data or secrets. Outputs are demos; review before using anywhere safety-critical or financial.

1. Quick image check

Upload a PNG or JPG and we run a lightweight browser vision model. Everything stays on-device; results are for learning, not for medical use.

Click or drop an image

2. Teach a custom classifier

Define a few labels, upload small images, and train a shallow classifier in your browser. We cap sizes to keep things responsive.

Step 1 - Labels

Step 2 - Upload & assign

Upload images for Class A

We cap to ~40 images to keep your browser happy.

No images yet.

Step 3 - Train

Training runs in your browser using a frozen backbone and a tiny head. Good for intuition, not production.

3. Embedding explorer

See how your images cluster in feature space. Hover to see thumbnails; click to highlight neighbours. This is a visual sketch for intuition, not production explainability.

Train the custom classifier or upload images above to view embeddings.

4. Recent Vision runs

Quick links to your latest image jobs. View details in Control Room.

View in Control Room

No Vision jobs yet. Train a custom classifier above to see it here.