A binary image classification on Kaggle Malaria Cell Images Dataset with FastAI library with an accuracy of approx 97%

The dataset was divided into two classes of infected and uninfected cell images of which we have to create a model which can classify from between two.

I used Transfer Learning method with Resnet34 weights to train with additional fully connected nn layers and again by unfreezing( to train my images with the whole architecture from the scratch) to get an accuracy of 97%

Initially I reduced the size of original pixel to half and used transfer learning with resnet on that cropped pixel dataset, after that I again used transfer learning method with cropped pixel trained model weights to train my new original sized images dataset which did the trick without overfitting

The dataset was obtained from kaggle dataset Website: https://www.kaggle.com/iarunava/cell-images-for-detecting-malaria

Anybody interested in playing with my model can use the above .pkl file. My whole procedure with detailed instruction is given in my above uploaded notebook checkout if interested.