✍️ BanglaOCR: A Platform to Collect Bangla Handwriting for OCR Models

Bangla OCR Handwriting Data is a full-stack web application designed to crowdsource handwritten Bangla text for building large-scale OCR datasets. It provides a clean submission portal for users and a robust backend for data moderation, tracking, and export.

---

🌐 Live Demo

• https://bangla-ocr-data-web.vercel.app/ (frontend)
• https://hossainimamrony71.pythonanywhere.com/ (API Endpoint)

---

🧠 Key Features

• Weighted random Bangla prompt selection
• Direct, secure client-side uploads to Cloudinary
• Real-time status updates during submission
• RESTful API using Django + DRF
• Admin-only moderation dashboard (via API for now)
• Verified submission export as .zip with CSV labels
• Decoupled architecture: Vanilla JS frontend + DRF backend

---

🔧 Backend: Django REST Framework

The backend powers secure APIs for serving prompts, logging submissions, moderating content, and exporting verified data.

📦 API Endpoints

GET /api/prompt/

• Purpose: Serve a Bangla text prompt using weighted random logic (prompts with fewer submissions appear more)
• Permissions: Public
• Response:

{
  "id": 1,
  "text": "বাংলা আমার মাতৃভাষা।"
}

---

GET /api/cloudinary-config/

• Purpose: Provide a secure signature for client-side Cloudinary uploads
• Permissions: Public
• Response:

{
  "api_key": "...",
  "cloud_name": "...",
  "timestamp": "...",
  "signature": "..."
}

---

POST /api/submissions/

• Purpose: Log metadata after successful Cloudinary upload
• Permissions: Public
• Body:

{
  "prompt": 1,
  "image_url": "https://res.cloudinary.com/.../image.jpg",
  "public_id": "bangla_ocr_handwriting_data/..."
}

---

GET /api/moderator/pending/

• Purpose: View unreviewed submissions for moderation
• Permissions: Admin only
• Returns: List of pending entries with image URLs and prompts

---

PUT /api/moderator/review/<int:id>/

• Purpose: Mark a submission as "verified" or "rejected", with optional notes
• Permissions: Admin only
• Example Update:

{
  "status": "verified",
  "notes": "Clear handwriting."
}

---

GET /api/download/verified-submissions/

• Purpose: Download a .zip with:

  • Folder of verified images
  • labels.csv (image → prompt mapping)

• Permissions: Admin only

---

🎨 Frontend: Vanilla JavaScript SPA

The frontend offers a minimal and intuitive experience for users contributing handwritten Bangla samples.

🔄 Flow Summary

1. Fetch Prompt

   • Calls /api/prompt/ and displays it using typewriter animation.

2. Select & Preview Image

   • Users upload an image which is locally previewed on-screen.

3. Upload & Submit

   • Fetch secure Cloudinary signature from /api/cloudinary-config/
   • Upload image directly to Cloudinary
   • Send metadata to /api/submissions/ for logging

4. Live Feedback

   • Users see real-time status updates: CONNECTING... → TRANSMITTING... → DONE ✅

---

🚀 Future: Admin Moderation Dashboard

Currently, moderation is done via API tools (like Postman). A future admin frontend is planned with:

Planned Features

• 🔐 Login-authenticated admin interface
• 📋 View & moderate pending submissions
• ✅ Approve/Reject entries with image preview
• 📥 One-click dataset export (.zip + labels.csv)
• 📊 Submission analytics dashboard
• 🛠 Built with React or Vue for responsiveness

---

📁 Dataset Export Format

verified_dataset.zip
├── images/
│   ├── user1_image.jpg
│   └── user2_image.jpg
└── labels.csv

# labels.csv format:
filename,text
user1_image.jpg,বাংলা আমার মাতৃভাষা।
user2_image.jpg,আমার সোনার বাংলা...

---

🛠️ Tech Stack

Layer	Technology
Backend	Django + DRF
Frontend	HTML, CSS, JS
Cloud	Cloudinary (uploads)
Auth	Django Admin
Dataset	.zip + CSV export

---

🙌 Contributing

Pull requests are welcome! If you plan on introducing breaking changes, please open an issue first to discuss the proposal.

---

📄 License

MIT License © RONY

---