A conceptual model to detect and verify signatures on bank cheques. This was my project in the "CSP 502: Computer Vision" course at Ahmedabad University.
We devised an OCR based approach to localize signatures along with Connected Components, Linesweep algorithm & geometric features with Support Vector Machine (SVM), Artificial Neural Networks (ANN) classification to verify the authenticity of signatures with a 91% hit rate.
Reference for connected components code: https://github.com/spwhitt/cclabel
Signature_Detection folder contains codes for different approaches: Contour features-based, OCR + Connected Components labelling, OCR + LineSweeping algorithm.
Signature_verification folder contains codes for SVM and ANN classification.
- Install dependencies via:
pip install -r requirements.txt - For execution of signature detection approaches
- OCR approach - run
python acheque.py - Connected Components approach - run
python cclabel.py - LineSweep algorithm, run
python lineSweepDetect.py - Contour feature based algorithm, run
python signature_detection.py
- For execution of signature verification approaches,
- For SVM based classification, run
python run.py - For ANN based classification, run
python test.py