Hugging Face's logo

Spaces:
jeffliulab
/
card-calibration-v1
Running

App Files Community
card-calibration-v1 / README.md
jeffliulab's picture
jeffliulab
Update Space README: link to dataset and improve metadata
518b7af verified
preview code
|
Raw History Blame Contribute Delete
1.98 kB

A newer version of the Gradio SDK is available: 6.30.0

Upgrade
metadata 
title: Card Calibration
emoji: 🎨
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.23.0
python_version: '3.10'
app_file: app.py
pinned: false
license: mit
models:
  - jeffliulab/card-calibration-v1
datasets:
  - jeffliulab/card-calibration-v1-data
tags:
  - color-calibration
  - yolo
  - computer-vision

Card Calibration β€” Live Demo

Upload a photo containing a color calibration card. The system detects the card, extracts color reference patches, and predicts the true color of the target patch under standard lighting.

How It Works

  1. Card Detection β€” YOLOv8 locates the calibration card in your photo
  2. Pattern Detection β€” A second YOLOv8 identifies 4 patches: red circle, green triangle, blue pentagon, black box (target)
  3. Feature Engineering β€” Extracts 12-D feature vector (9 reference deltas + 3 target RGB)
  4. Color Prediction β€” XGBoost or Random Forest predicts the true RGB under standard lighting

Best result: XGBoost (Bayesian-tuned) β€” Lab Mean Ξ”E = 4.59 (commercial printing standard)

Models

File Description
yolo_first.pt YOLOv8-nano β€” card detection
yolo_second.pt YOLOv8-nano β€” pattern detection
xgboost_v1.pkl XGBoost calibration (best, Ξ”E=4.59)
random_forest_v1.pkl Random Forest calibration

Models are hosted on jeffliulab/card-calibration-v1 and downloaded automatically on first request via huggingface_hub.

Training Data

255 hand-collected photos, augmented to 2,294 samples. Full dataset available on jeffliulab/card-calibration-v1-data.

Links