Camera evidence · experimental
Look at the evidence.
Then check the status.
Read OPEN/CLOSED text, detect supported common objects, and measure brightness. Compare these observations with owner-entered hours. Darkness and no detected people alone do not prove a store is closed.
1. Source and models
Models download from their public CDNs. Image processing runs on this device; frames are not uploaded. Initial loading needs internet. Supported objects are limited to COCO-SSD classes, not every object or a trained shutter detector.
Models not loaded.
Use a close, clear sign image. OCR may read advertisements or background text; review its output before relying on it.
2. Schedule and observations
Hours are configured by the operator, not extracted from a photo. Overnight opening hours are supported. This schedule is saved on this device.
Load models and select an image or webcam.
- Schedule expectation
- Not checked
- Sign text
- Not read
- Lighting
- Not measured
- People
- Not detected yet
- Objects
- Not detected yet
3. Send observations to your camera board
Use a camera board from Operator studio. Three matching, acknowledged observations ten seconds apart confirm a public label. Pairing authenticates the sender; it cannot independently prove the camera evidence is true.
Observations stay on this page until sending is enabled.
What this does and does not detect
Object detection draws boxes for supported common objects, including people. OCR reads visible English text; brightness describes the selected frame, not the store’s electrical lights. A person can be out of view and a camera can adjust exposure. This experimental route does not ingest RTSP CCTV or combine two feeds. Do not use it as a guarantee of service.