Deep Learning Code Base for Tranining Models For Different Applications

This Repo Consists of Deep Learning Code Base for Creating Custom Models, Data loaders, losses, down/up sampling training and Data Augmentations techiniques using PYTorch as the main framework for training, It is used to train models for Dectecting Diabetic retinopathy of fundas Images.: https://morning-anchorage-56517.herokuapp.com/adl.html

Folder Structure:

Models Package for Developing Different Models
Losses For Developing Different Loss functions
dataloaders Customised Data Loaders
Custom Transformations Developing custom Image Transformations for Data Augmentation on Fly
asved_models To save the models strcture and its weights (Trainned)
utility Code for Helper funtions Eg: Visualization of Data.

Infrastructure:

Python 3.6
pytorch
Miniconda(used to create virtual envinorments).
"requirments.txt" Consists of all the dependices

Deployed Model on Heroku using ExpressJs, NodeJS as an API

https://morning-anchorage-56517.herokuapp.com/adl.html