Featured Project
OCR Image Quality Evaluation Tool
A Python tool for evaluating image quality metrics that directly impact OCR accuracy — built for government and research workflows that need reliable text extraction from scanned documents and imagery. Provides CLI and PySide6 GUI interfaces for single-image analysis, batch evaluation, and ground-truth processing (including Label Studio exports). Detects text regions with bounding boxes, compares results against ground truth, and reports character/word accuracy, precision, recall, and F1 scores alongside sharpness, contrast, noise, and SSIM metrics — with CUDA-accelerated processing and exportable visualizations.
- Python
- PyTorch
- OpenCV
- PySide6
- EasyOCR
- Tesseract