✓ EVIDENCE-FIRST CARD INTELLIGENCE

See the evidence before you submit.

Photographic pre-grading for trading cards with exact-card identification, centering, capture-quality gates, and reference-relative forensic evidence.

Open guided capture →

◎32 regionsstandardized forensic map

✓Abstention firstrecapture when evidence is weak

◇Traceableversioned evidence chain

Small bun. A closer look at what you love.Upload → inspect → evidence → stamp.

LEFT / RIGHT49.0% / 51.0%

TOP / BOTTOM50.0% / 50.0%

REFERENCE

✦

CARD PREVIEWExact-reference centering

Exact-reference centering previewVisual measurement first, uncertainty shown when reference alignment is weak.

POKÉMONSPORTSMAGICYU-GI-OH!ONE PIECELORCANAMore card families planned

01 · CAPTURE

Upload both sides

✓ Photos are analyzed as evidence, not treated as ground truth.

Use a flat, unsleeved card when possible. Keep the camera square to the card, use diffuse light, avoid heavy glare, and preserve full resolution.

↑FRONT of card

Drop a photo here or click to browse

JPG, PNG or WebP · up to 20 MB

FRONT

↑BACK of card

Drop a photo here or click to browse

JPG, PNG or WebP · up to 20 MB

BACK

▣Square-onKeep all four card corners visible.

✦Diffuse lightAvoid sleeve haze and blown highlights.

◎High resolutionDo not crop away physical card edges.

ANALYSIS COMPLETE

Your evidence report

IDENTIFIED CARDIdentity not establishedCardGrade abstains if exact-match evidence is insufficient.

REFERENCE MATCH—Geometric verification required

FORENSIC CAPTURENOT RUNExact reference required

LEFT / RIGHT—

TOP / BOTTOM—

EXACT-REFERENCE FORENSICS

Where the photograph differs

not run

A verified local canonical reference is required before exact-reference evidence can run.

ALIGNMENT—

NULL PROFILE—

MATERIALunknown

TOP OBSERVATIONS

Evidence, not guessed defect labels

No exact-reference observations are available for this scan.

ALPHA FEEDBACK

Was this result useful?

Tell us how the experience felt. Feedback improves the product and camera coach; it never changes this card's grade.

RatingResult useful?Use again?

Optional note

02 · CENTERING STUDIO

Make centering understandable at a glance

REFERENCE-RELATIVE

CardGrade presents left/right and top/bottom balance visually. When an exact card is identified, the measurement can be anchored to a matched canonical reference rather than one universal border guess.

01Physical-card framePerspective correction establishes the cut boundary.

02Printed-reference alignmentInterior artwork/features align the exact printing without using the outer cut edge.

03Confidence + abstentionBad reference geometry or weak alignment yields uncertainty, not false precision.

LEFT / RIGHT49.0% / 51.0%

TOP / BOTTOM50.0% / 50.0%

REFERENCE

✦

CARD PREVIEWExact-reference centering

03 · TECHNOLOGY

A grade should have an evidence trail

01⌕

Identify

Find the exact printing using retrieval, OCR hints, and geometric verification.

02⌗

Align

Rectify the capture and align a clean reference using robust homography and precision refinement.

03◎

Center

Estimate print/cut offset from reference-relative geometry instead of one generic border heuristic.

04⌜

Inspect

Score corners and edges using standardized forensic regions and capture-quality controls.

05≋

Compare

Build a reference-relative anomaly map while suppressing glare, illumination changes, and registration noise.

06✓

Calibrate

Compare evidence to a benign null and request recapture when photographic quality is insufficient.

CERTIFICATE LOOKUP

Verify a CardGrade analysis

Prototype scan IDs fingerprint the submitted image pair. Permanent physical-card identity across different photographs is a later milestone.

⌕

PRIVATE ALPHA ACCOUNT

Keep your collection across devices

SIGNED OUT

Accounts are optional during private alpha. Passwords are memory-hard hashed, sessions use secure cookies, and account data can be exported or deleted.

EmailPasswordI agree to participate in this experimental private alpha and understand CardGrade results are research/pre-grading evidence, not an official grading-company grade.

ACCOUNTNot signed inPrivate alpha · optional

Export returns your profile, collection, feedback, capture sessions, and a list of your opted-in research photos with authenticated download links. You can delete research photos without deleting your account.

ALPHA COLLECTION

Your recent CardGrade scans

Signed-in alpha testers use the account collection across devices. Signed-out testers keep a browser-local fallback collection.

No cards saved on this device yet.

RESEARCH TRANSPARENCY

We show what is measured—and what is not.

Implemented capability, synthetic engineering evidence, and real-world validation remain separate states.

FORENSIC ENGINEv0.16No numerical grade effect.

DINOv3CHECK STATUSOfficial checkpoint/runtime status.

REAL DEFECT CHECKPOINTNOT PROMOTEDReal rights-cleared labels still required.

PUBLIC BETABUILDINGFrontend + production infrastructure in progress.

PUBLIC BETA ROADMAP

What we are building next

Public product work and scientific validation can progress in parallel without inflating model claims.

01Production account + authentication layer

02Object storage + Postgres migration

03Real rights-cleared capture program

04DINOv3 measured reference retrieval

05Real defect checkpoint + frozen test

06Shadow-mode validation

07Billing / collection / certificate pages

08Public beta + monitoring