Welcome to the Art Signature Authentication project repository! This cutting-edge project utilizes Machine Learning to accurately verify artist signatures, addressing long-standing challenges in the art market. 🖌️✨
The authenticity of an artist's signature plays a pivotal role in determining the value and legitimacy of artworks. Traditional methods, reliant on subjective expert opinions, are prone to errors and inefficiencies. Our project introduces a machine learning-based solution to bring higher accuracy, objectivity, and scalability to signature verification.
🔍 Challenges in the Art Market:
- Rampant forgeries and misattributed artworks lead to financial losses and diminished trust.
- Traditional verification methods are costly, subjective, and error-prone.
🌟 Our Solution: An automated, reliable machine learning model that:
- Reduces dependence on manual processes.
- Increases precision and scalability for the art community.
- Data Augmentation: Expanded a dataset of 576 signature images to 9,080 images using techniques like rotation, zoom, pixelation, and more.
- Preprocessing: Enhanced features with thresholding, centering, and dimensionality reduction using SVD.
- Model Comparisons: Evaluated four machine learning models, including Logistic Regression, Random Forest, Support Vector Machines (SVM), and Convolutional Neural Networks (CNN).
- Superior Model: Achieved a remarkable 99.46% accuracy with CNN, demonstrating its capability to handle complex signature datasets.
- Signature Analysis: Accurately classifies artist signatures.
- Forgery Detection: Identifies subtle differences to detect forgeries.
- Scalability: Accommodates additional artists' signatures effortlessly.
- User-Focused Design: Simplifies authentication for collectors, galleries, and auction houses.
notebooks/: Jupyter Notebooks for model development.data/: Processed and augmented datasets.
| Model | Accuracy | Precision | Recall |
|---|---|---|---|
| Logistic Regression | 73.95% | 50%-86% | Varies |
| Random Forest | 88.71% | 74%-96% | Strong |
| Support Vector Machines | 81.11% | 42%-100% | Good |
| Convolutional Neural Network | 99.46% | Near Perfect | Excellent |
This project provides:
- Enhanced confidence for collectors and art institutions.
- Faster and more cost-effective authentication.
- A powerful tool to combat art forgery, fostering greater trust in the art community.
Have questions or suggestions? Feel free to contact me at gauravadavkar13@gmail.com
Thank you for exploring the Art Signature Authentication project! Let’s revolutionize the art world together. 🎉