sataseyu-AI-verification / QUICKSTART.md
Anurag Banerjee
Model added
3a32bd4
|
Raw
History Blame Contribute Delete
4.24 kB
# ๐Ÿš€ Quick Start Guide
Get the Certificate Verification System running in **5 minutes**!
---
## โšก Fastest Way to Run
### **Option 1: Demo Mode (No Setup Required)**
```bash
# Clone repository
git clone https://github.com/YourUsername/certificate-verifier.git
cd certificate-verifier
# Install dependencies
pip install -r requirements.txt
# Run app
streamlit run main.py
```
**That's it!** App runs in demo mode - no API keys needed.
---
## ๐Ÿ”ง Full Setup (With OCR API)
### **1. Install Dependencies**
```bash
pip install -r requirements.txt
```
### **2. Get Free OCR API Key**
1. Visit: https://ocr.space/ocrapi
2. Sign up (free)
3. Copy your API key
### **3. Configure Environment**
Create `.env` file:
```bash
OCRSPACE_API_KEY=your_api_key_here
```
### **4. Initialize Database** (Optional)
```bash
python init_db.py
```
### **5. Run Application**
```bash
streamlit run main.py
```
### **6. Open Browser**
```
http://localhost:8501
```
---
## ๐Ÿ“ฑ Usage
### **Basic Workflow:**
1. **Upload** certificate image (JPG/PNG)
2. **Click** "Verify Certificate"
3. **View** results with confidence scores
4. **Download** verification report (JSON)
### **Demo Mode Testing:**
1. Enable "Demo Mode" in sidebar
2. Upload any certificate image
3. System uses sample OCR data
4. See how verification works!
---
## ๐ŸŽฏ What Happens on First Run?
### **Automatic Downloads:**
1. **YOLOv8 Model** (~6 MB)
- Downloads from Hugging Face
- Takes ~10 seconds
- Cached for future runs
2. **ViT Model** (~1 GB) (Only if seal verification enabled)
- Downloads from Hugging Face
- Takes ~5 minutes (depending on bandwidth)
- Cached for future runs
**After first run:** Everything loads instantly from cache!
---
## โœ… Verification Steps
The system performs **3-layer verification**:
### **Layer 1: OCR Text Verification**
- Extracts text from certificate
- Finds registration number
- Matches against database
- Calculates confidence score
### **Layer 2: YOLOv8 Seal Detection**
- Detects seals/stamps in image
- 99% detection accuracy
- Returns bounding boxes
### **Layer 3: ViT Seal Classification**
- Classifies each seal as Real/Fake
- Uses Vision Transformer AI
- Provides confidence scores
### **Final Decision:**
- Combines all layers
- Security-first logic
- High-confidence fake โ†’ Rejection
---
## ๐Ÿ” Test Certificates
The database includes sample certificates you can test:
**Sample Registration Numbers:**
- `ABC2023001` - Saksham Sharma, DevLabs Institute
- `ABC2022007` - Prisha Verma, Global Tech University
- `1BG19CS100` - Vikram Verma, VTU (from demo mode)
---
## ๐ŸŽฎ Features to Try
### **In Sidebar:**
โœ… **Demo Mode** - Test without API keys
โœ… **Seal Verification** - Enable AI seal detection
โœ… **OCR Language** - Select certificate language
โœ… **System Status** - Check all components
### **After Verification:**
โœ… **Detailed Results** - Step-by-step breakdown
โœ… **Confidence Scores** - For each verification layer
โœ… **Download Report** - JSON export
โœ… **Detected Seals** - View cropped seal images
---
## ๐Ÿšจ Troubleshooting
### **Issue: ModuleNotFoundError**
```bash
pip install -r requirements.txt
```
### **Issue: OCR API Error**
- Enable "Demo Mode" in sidebar, or
- Check API key in `.env` file
### **Issue: Models not downloading**
- Check internet connection
- Models download automatically on first run
- Look for download progress in terminal
### **Issue: Database error**
```bash
python init_db.py
```
---
## ๐Ÿ“š Next Steps
### **Deploy to Cloud:**
See [DEPLOYMENT.md](DEPLOYMENT.md) for Streamlit Cloud deployment
### **Customize:**
- Edit `certs.db` to add your certificates
- Modify verification thresholds in `verifier.py`
- Add custom regex patterns for registration numbers
### **Integrate:**
- Use as Python library
- Build REST API wrapper
- Integrate with existing systems
---
## ๐Ÿค Need Help?
- **Documentation:** [README.md](README.md)
- **Deployment:** [DEPLOYMENT.md](DEPLOYMENT.md)
- **Issues:** Open on GitHub
- **Questions:** Contact maintainer
---
## ๐ŸŽ‰ You're Ready!
**Start verifying certificates with AI-powered accuracy!**
```bash
streamlit run main.py
```
Happy verifying! ๐ŸŽ“โœจ