https://github.com/efecanxrd/signature-recognition

Signature recognition using template matching and correlation
https://github.com/efecanxrd/signature-recognition

opencv signature signature-detection signature-recognition signature-validation signature-verification signatures template-matching

Last synced: 3 months ago
JSON representation

Signature recognition using template matching and correlation

• Host: GitHub
• URL: https://github.com/efecanxrd/signature-recognition
• Owner: efecanxrd
• License: mit
• Created: 2023-08-09T18:14:29.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-08-11T11:05:16.000Z (almost 2 years ago)
• Last Synced: 2025-01-15T11:50:00.098Z (5 months ago)
• Topics: opencv, signature, signature-detection, signature-recognition, signature-validation, signature-verification, signatures, template-matching
• Language: Python
• Homepage:
• Size: 147 KB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
 Signature Recognition 

 Signature recognition using template matching and correlation 



 Setup The Project 

 


 Download Dataset 


 https://www.kaggle.com/datasets/robinreni/signature-verification-dataset 



 Install the libraries with pip 

 
pip install tkinter opencv-python pandas numpy Pillow 

 How this is working? 

  
Uses the template matching technique in OpenCV to match the signature in an image with the signatures from the dataset.

First loads the image and converts it to grayscale using the cv2.imread and cv2.cvtColor functions. Next, we set a correlation value for template matching.


As searches for the signature, it iterates over the folders in the dataset and loads each signature image. The code implements template matching using the cv2.matchTemplate function and checks if a match is found using the correlation value. 


This function shifts the template image over the source image and calculates a similarity measure between the template and each region of the source image. The similarity measure used in this case is cv2.TM_CCOEFF_NORMED.

The result of template matching is a 2D array of similarity values. The code uses the np.where function to find all locations in this sequence where the similarity value is greater than or equal to the correlation value.