Exploration of CNN's in Blood Cell Classification

Problem

The task that this project attempted to tackle was the classification of blood cells. We wanted to explore and compare different neural networks and their ability to classify these images from a Kaggle dataset. The relevance of this project was to attempt to contribute a novel classification method that can aid medical professionals in making more informed diagnoses.

Solution

We first choose a well-known model as our baseline to which we would compare the rest of the models. In this case, the VGG16 was chosen to serve as the base. Then we tested 4 different models and compared their results with the baseline. After these results, we attempted to implement our findings into a new model that had various aspects of the previous 4. Once we implemented this new model, we again tested and compared it with our baseline to see if our rationale would hold up in testing.