#5 [Refund the project architecture] Enhancement toward project completion

[HugoHakem]

Refund the whole project architecture

Ideally we'd like:

1. folder for image
2. a folder for dataset retrieval:

• .ipynb where the selection of the JUMP dataset is occurring. This shall include mAP computation
• .py for plotting
• .py balancing dataset
• .py for image retrieval into crops etc.

3. Another folder where the analysis is occurring:

• .py for data splitting
• .py for custom dataset
• .py for classifier architecture
• .py for classifier lightning architecture training
• .py for classifier training
• .py for classifier validation metrics
• .py for generative architecture
• .py for generative lightning architecture training
• .py for generative training
• .py for generative validation metrics
• .py for captum attribution methods
• .py for DAC computation
• .py for mask retrieval and dataset creation
• .py for MARIMO interface

4. parameter file that decide the color of each channel, the number of fold, the seed, which fold we are playing with, the classifier chosen, the generative model chosen, the amound of GPU, the hyperparameter tuning, the generation mode (ref or lat) etc...
   Everything should be referring to that parameter file so that there is no any problem of consistency across script.

Maybe a smarter architecture could be imagined. Suggestions are welcomed.

[HugoHakem]

Refactoring lunched, see refactoring branch.
Goal: getting back to the project by cleaning up everything and keeping track of the reasoning.
Everything should be written in a python package fashion.