This project is a white blood cell classification project that utilizes the pytorch framework to classify white blood cell images as eosinophil, lymphocyte, monocyte, or neutrophil.
The firs 4CF0 t file, newmodel.py, is an attempt at training a machine learning model from scratch. It has some features that can be enabled, such as gaussian or high pass image filters, to possibly help with cell type classification. With a single epoch of training, the model achieved 53% accuracy.
The second file, pretrained.py, leverages the pretrained shufflenet model, and fits the model for the needs of image classification. This model, also on a single training epoch, achieved a 91% accuracy on the testing dataset.
Each of these files has a training method, which trains the given model into a .pth file, and a testing function, which useses the test dataset to assess the accuracy of a given model file.
To run newmodel.py without high pass filter, go into MyNetwork() and comment out the high pass function. Running the code will save a .pth file on the computer which represents the model that was just trained.