The EDA code can be utilized to clean the data and create a model architecture with custom CNN model training (Also contains code for a VGG inspired model)
Ths submission code can be used to create a model using the same weights as the trained model and provide inference.
There might be some irregular comments from the code pieces I borrowed from the Kaggle Repositories or Some of the ones I had created but changed the architecture later. Please do not Mind those I will work on cleaning and republishing the notebooks (or somebody else who might be interested in doing that can also thake it up :) )