Design model architecture

[DPS100]

Figure out type and parameters of model to use for each dataset.

• [x] See if we need to do any preprocessing with data from real or synthetic data set
• [x] Figure out what kind of model to use (TensorFlow CNN?)
• [ ] Train, tune, and find best hyperparameters for a model on each true dataset

[Bashir1999]

Deep Preprocessing was not needed. PyTorch can be used to implement CNN cz of ease of use. I have added the notebook for training models. Take a look at the code and comment if you find issues. The dataset is organized into files: train -> is a folder, test -> folder, syn_img -> folder, inside each folder, there are three other folders each containing one class, apple folder, banana folder, and orange folder