HUGGINGFACE_CHECKLIST.md

📦 Hugging Face Deployment - File Checklist

✅ Files Created for Deployment

Core Application Files

• ✅ app.py - Gradio interface with OCR processing
• ✅ requirements.txt - Python dependencies (Gradio + eDOCr2)
• ✅ packages.txt - System dependencies (Tesseract, Poppler)
• ✅ README.md - Space description with YAML frontmatter
• ✅ .gitattributes - Git LFS configuration for model files

Documentation

• ✅ DEPLOYMENT.md - Complete deployment guide
• ✅ run_local.bat - Windows quick start script
• ✅ run_local.sh - Linux/Mac quick start script

Required Folders

• ✅ edocr2/ - Main package (already exists)

  • ✅ edocr2/tools/ - OCR pipelines
  • ✅ edocr2/keras_ocr/ - OCR models
  • ⚠️ edocr2/models/ - Model files (MUST DOWNLOAD)

• ✅ tests/test_samples/ - Example drawings (optional)

🔴 IMPORTANT: Download Model Files

Before deploying, download these 4 files and place in edocr2/models/:

1. recognizer_gdts.keras (67.2 MB)
2. recognizer_gdts.txt (85 bytes)
3. recognizer_dimensions_2.keras (67.2 MB)
4. recognizer_dimensions_2.txt (42 bytes)

Download from: https://github.com/javvi51/edocr2/releases/tag/v1.0.0

📋 Pre-Deployment Checklist

Local Testing

• Models downloaded and placed in edocr2/models/
• Run python app.py locally
• Test with sample images
• Verify all outputs (image, JSON, ZIP)

Hugging Face Setup

• Hugging Face account created
• Git LFS installed
• New Space created on Hugging Face

File Verification

• All files present in folder
• Model files in correct location
• .gitattributes configured for LFS
• README.md has YAML frontmatter

🚀 Deployment Steps

1. Download Models

Windows PowerShell:

cd edocr2-main
New-Item -ItemType Directory -Force -Path edocr2\models
cd edocr2\models

Invoke-WebRequest -Uri "https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_gdts.keras" -OutFile "recognizer_gdts.keras"
Invoke-WebRequest -Uri "https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_gdts.txt" -OutFile "recognizer_gdts.txt"
Invoke-WebRequest -Uri "https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_dimensions_2.keras" -OutFile "recognizer_dimensions_2.keras"
Invoke-WebRequest -Uri "https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_dimensions_2.txt" -OutFile "recognizer_dimensions_2.txt"

cd ..\..

Linux/Mac:

cd edocr2-main
mkdir -p edocr2/models
cd edocr2/models

wget https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_gdts.keras
wget https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_gdts.txt
wget https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_dimensions_2.keras
wget https://github.com/javvi51/edocr2/releases/download/v1.0.0/recognizer_dimensions_2.txt

cd ../..

2. Test Locally (Optional but Recommended)

Windows:

run_local.bat

Linux/Mac:

chmod +x run_local.sh
./run_local.sh

Open: http://localhost:7860

3. Create Hugging Face Space

1. Go to https://huggingface.co/spaces
2. Click "Create new Space"
3. Settings:
   • Name: edocr2 (or your choice)
   • License: MIT
   • SDK: Gradio
   • Hardware: CPU Basic (free)

4. Clone Space Repository

git clone https://huggingface.co/spaces/YOUR_USERNAME/edocr2
cd edocr2

5. Copy Files

Windows:

xcopy /E /I C:\path\to\edocr2-main\* .

Linux/Mac:

cp -r /path/to/edocr2-main/* .

6. Setup Git LFS

git lfs install
git lfs track "*.keras"
git add .gitattributes

7. Commit and Push

git add .
git commit -m "Initial deployment of eDOCr2"
git push origin main

Note: Upload may take 5-10 minutes for large model files.

8. Wait for Build

• Go to your Space URL
• Wait 5-10 minutes for build
• Check "Logs" tab for errors

✅ Verification

Once deployed:

• Space shows Gradio interface
• Models load successfully (check logs)
• Can upload images
• Processing works
• Results display correctly
• Download ZIP works

🎯 Your Space URL

After deployment, your Space will be at:

https://huggingface.co/spaces/YOUR_USERNAME/edocr2

📊 Expected Performance

CPU Basic (Free)

• Processing time: 20-30 seconds per image
• Memory: 2 GB RAM
• Cost: FREE

T4 GPU (Paid)

• Processing time: 5-10 seconds per image
• Memory: 16 GB RAM
• Cost: $0.60/hour

🐛 Common Issues

"Models not found"

• Ensure models are in edocr2/models/
• Check Git LFS tracked the files
• Verify file names are correct

"Out of memory"

• Upgrade to GPU hardware
• Or reduce max_img_size in app.py

"Build failed"

• Check logs for specific error
• Verify all dependencies in requirements.txt
• Ensure packages.txt has system deps

📚 Resources

• Deployment Guide: See DEPLOYMENT.md
• Hugging Face Docs: https://huggingface.co/docs/hub/spaces
• Gradio Docs: https://gradio.app/docs
• Original Repo: https://github.com/javvi51/edocr2

🎉 Success!

Once deployed, share your Space:

🔗 https://huggingface.co/spaces/YOUR_USERNAME/edocr2

---

Questions? Check DEPLOYMENT.md for detailed troubleshooting.