Camera guidance
Help a person frame the right object and collect enough detail before leaving the camera.
Skip to content01 02 03 04
GRADE OPTIC VISUAL ARCHITECTURE
A unified intelligence pipeline from lens to decision.
Grade Optic treats camera control, object geometry, recognition, evidence qualification, and the human handoff as one coordinated system rather than a collection of disconnected model calls.
LIVE CAPTUREBOUNDARYIDENTITYVISUAL SIGNALSQUALIFIED OUTPUT
FROM LENS TO DECISION
Five handoffs. One source photograph.
The hard engineering is not a single model call. It is keeping camera state, geometry, image quality, identity, and product rules aligned from the first live frame to the report a person opens.
- 01
Guide the camera
The browser reads a live stream and looks for a usable object boundary, manageable motion, focus, and light. It can guide another shot before weak pixels become an expensive result.
- 02
Preserve the source
The submitted photograph remains the source record. Crops, aligned views, thumbnails, and display art are derived representations with different jobs.
- 03
Find the physical object
Boundary models separate a card from the mat, sleeve, fingers, glare, and surrounding room. Geometry is mapped back onto the saved still so the system does not confuse preview coordinates with sensor pixels.
- 04
Read in layers
Identity, visible condition signals, and market context are separate questions. Each layer can succeed, fail, or ask for another view without forcing the others to pretend.
- 05
Qualify the result
A report is useful only when its inputs and limits travel with it. Missing backs, blur, glare, and unsupported conclusions stay visible in the output.
WHAT ACTUALLY TRANSFERS
The subject changes. The capture problems repeat.
Object geometry
Find boundaries, correct perspective, and carry coordinates from the preview into the saved image.
Layered analysis
Keep recognition, visual observations, and outside context separate so one weak layer does not contaminate another.
Human handoff
Show the source, the useful signal, and the limitation at the moment someone needs to choose what happens next.
Go Precheck is the flagship implementation of this architecture. The underlying discipline remains constant across the Grade Optic portfolio: control the input, preserve provenance, separate analysis layers, and deliver a qualified result.