PikaFam Project Room

Access code required

This page has a lightweight access gate so it does not get opened casually from the projects list.

*** BEST VIEWED WITH CURIOSITY ENABLED ***
Centering Intelligence
TCGMarketPulse + PulseLine
Mini-site
Overview 01 Implementation 02 OpenClaw instance 03 Skills 04 Architecture 05 Roadmap 06
Scope: implementation planning for centering eligibility. This does not claim final grading outcomes.
NemoLink / Projects / TCG Centering / Build plan
Draft spec updated Aug 20, 2026

Build plan

Implementation specifics

Concrete engineering requirements for turning card photos into measured centering ratios, overlays, grader eligibility, and product-ready reports.

Worker stackDB tablesAPI contracts
Worker stack
Python + CV
OpenCV first; segmentation model when needed.
Artifact store
images + JSON
Original, normalized crop, overlay, measurements.
Operator path
manual handles
Needed for low-confidence boundaries.
API contract
/analyze
Async job with status/result endpoints.

Concrete implementation components

What has to be built, beyond the user-facing page.

Image intake endpointAccept front and optional back image, enforce size/type limits, strip metadata, assign analysis id, and enqueue processing.
Quality preflightCompute blur, glare, crop, resolution, corner visibility, background contrast, and sleeve/slab indicators.
Card boundary detectorFind outer quadrilateral, refine corners, reject cropped/uncertain images, and calculate a transform matrix.
Perspective normalizerRectify to canonical card aspect ratio. Store normalized image for repeatable measurement and debugging.
Layout classifierPick measurement strategy for vintage, modern bordered, full art, trainer, Japanese, reverse, and card back layouts.
Inner-boundary detectorDetect artwork/print window or relevant border landmarks. Provide manual override coordinates when confidence is low.
Rules comparison engineCompare ratios to versioned grader profiles with uncertainty-aware pass/borderline/fail states.
Report rendererGenerate JSON result, concise PulseLine text, detailed TCGMarketPulse panel, and annotated overlay.

Minimum repository additions

Suggested folder structure.

tcg-centering/
  worker/
    analyze.py
    quality.py
    detect_card.py
    rectify.py
    classify_layout.py
    measure.py
    render_overlay.py
  rules/
    psa.json
    cgc.json
    bgs.json
    sjc-placeholder.json
  api/
    routes.ts or routes.py
  migrations/
    centering_analysis.sql
  eval/
    golden_set_manifest.json
    run_eval.py

Database tables to add

Enough structure for auditability and later model improvement.

TablePurposeKey fields
centering_analysisOne user-facing analysis jobid, user_id, card_id, status, created_at, completed_at, model_version, rules_version
centering_imageFront/back image artifactsanalysis_id, side, original_uri, normalized_uri, overlay_uri, width, height, sha256
centering_qualityPhoto preflight resultsblur_score, glare_score, crop_status, corner_visibility, warnings_json
centering_measurementMeasured ratios and boundariesside, left_px, right_px, top_px, bottom_px, lr_ratio, tb_ratio, uncertainty_pct, confidence
grader_rule_profileVersioned centering thresholdsgrader, grade_label, front_rule_json, back_rule_json, source_note, effective_date
centering_feedbackHuman correction and returned-grade loopanalysis_id, corrected_boundaries_json, returned_grader, returned_grade, notes
©2026 - NemoLink by Nemo Prime - Centering intelligence brief