project / 2025
Machine learning / computer vision
KU ID Verifier
An ID verification system that combines visual classification, dual OCR, and fraud checks in a usable workflow.
- built with Python / PyTorch / Streamlit / EasyOCR / Tesseract
- source View source ↗
notes on the build
Snapshot
KU ID Verifier focuses on a practical verification problem: how to validate institutional ID cards with enough confidence that the output feels useful rather than decorative.
The project layers visual classification, OCR extraction, fuzzy validation, and fraud checks into one pipeline, then exposes the result through a Streamlit interface that keeps the process legible.
Why it earned its place
It earns its place because it treats machine learning as one part of a real verification flow, not the whole story.
what mattered
- Dual OCR pipeline using EasyOCR and Tesseract for validation.
- Fraud detection layers including image hashing, manipulation checks, and duplicate prevention.
- Hardware-aware runtime path for Apple Silicon, CUDA, and CPU fallback.
next project