Offline Signature Verification System using Convolutional Neural Networks

The paper titled “Offline Signature Verification System Using Convolutional Neural Networks (CNN)” is developed to automate the process of verifying handwritten signatures, ensuring accuracy and security in authentication systems. This system was developed using Python, with Tensor Flow / Keras for deep learning model development and OpenCV for image preprocessing. The implementation integrates Canny Edge Detection, Gaussian Blur and Grayscale conversion to pre-process signature images, enhancing feature extraction for accurate classification. The system is designed to replace traditional manual verification methods, reducing human intervention and errors while improving efficiency. It includes functionalities to load signature images, pre-process them and classify signatures as genuine or forged based on deep learning predictions. Verification reports can be generated based on different test cases and accuracy metrics. Recent advances in deep learning, particularly CNN, provide a more powerful approach by enabling automatic learning of complex patterns from raw data and improving the accuracy of the model. This paper explores the application of CNNs to verify handwritten signatures through static signature images, addressing the inherent challenges posed by variability in writing styles and forgeries. By employing CNN-based feature extraction and classification, this work aims to improve the reliability and efficiency of signature verification systems, making them more effective for real-world applications. The proposed model achieves high accuracy, demonstrating its potential for practical implementation in secure authentication systems.