What is fingerprint-based-blood-group-detection-using--vision-transformer-and--ensemble-learning?

This repository contains a research project that tries to detect a person's blood group from a fingerprint image using machine learning. The idea is that fingerprint patterns may carry information that correlates with blood type, and the project tests whether image recognition models can learn that correlation. The project trains and compares four different image classification architectures: ResNet, VGG16, AlexNet, and LeNet. These are all established approaches for recognizing patterns in images. The dataset contains roughly six thousand to seven thousand fingerprint images sorted into eight folders, one for each blood group category (A+, A-, B+, B-, AB+, AB-, O+, O-). Each model is trained on those images and then evaluated on a test set. The code is organized as a set of Jupyter notebooks, one per model. A Jupyter notebook is an interactive document that mixes code and output so you can run one step at a time and see results inline. Each notebook loads the dataset, trains the model, and evaluates its accuracy. After training, accuracy and loss graphs are saved to a graphs folder, and bar charts comparing performance across all four models are saved to a performance metrics folder. To run any of the notebooks you need Python 3.8 or newer, PyTorch (a machine learning library), and a few other libraries listed in a requirements file included in the repository. You install them with a single command and then open the notebook of your choice. The README mentions a research paper as the foundation for the work but does not link to it directly. The project is released under the MIT license. No live demo or deployed application is described, this is a research and experimentation codebase.