QR Code Authentication Model

QR Code Authentication: Detecting Original vs. Counterfeit Prints 🤓 Developed a hybrid ML model combining traditional computer vision (HOG + SVM) with deep learning (CNN) to distinguish genuine from counterfeit QR codes. Utilized TensorFlow 2.10, scikit-learn, OpenCV, and scikit-image for image preprocessing, augmentation, and feature extraction😵‍💫. The CNN achieved 95.54% validation accuracy compared to 74% for the SVM, demonstrating resilience against scanning artifacts. 👀 Learned to integrate classical methods with deep learning and leverage GPU acceleration (where CUDA and cuDNN 😶‍🌫️ bother) for efficient model training.

This project detects original (first print) versus counterfeit (second print) QR code images using two approaches:

1. A traditional computer vision pipeline (HOG feature extraction with SVM).
2. A deep learning-based approach (CNN with data augmentation).

Project Structure

• dataset/
  Contains First_Print (originals) and Second_Print (counterfeits).

• notebooks/
  Contains interactive notebooks for data exploration (data_exploration.ipynb) and CNN training/visualization (cnn_training.ipynb).

• src/
  Contains Python scripts:

  • traditional_pipeline.py: SVM training using HOG features.
  • cnn_model.py: CNN model training.
  • utils.py: Utility functions for data loading and visualization.

• Results/

• requirements.txt
  Lists the required Python packages.

Setup and Execution

1. Create and activate a Python virtual environment.
2. Install the required packages using: requirements.txt (compatable with supported version)