will update it later
Problem Statement: This is an open ended project to provide DL based solutions.
Task: Develop a DL based model to solve the COVID19Action-Radiology-CXR https://ieee-dataport.org/open-access/covid19action-radiology-cxr
Submit: Notebook with executed solution on Pytorch, clearly indicating through comments and explained text/header in the .ipynb, detailing
have you taken any methods to compensate for class imbalance?
any conditions of hot restart during training?
present the training/validation loss curves and indicate how you have chosen the stopping criteria?
calculate the model space and inference compute complexity and indicate it on your .ipynb
the .ipynb should be self contained with any custom function defined in-scope, and the training function as well, and should take in a single folder with images at input, without any need to manually segregate them into train, test, validate sets
Executed .ipynb should necessarily be submitted for qualifying for evaluation
DO NOT SUBMIT trained model.
Additional Grades for Hosting the Model/Inference code on GitHub, writing a whitepaper on your method and posting it on ArXiv, and providing the link with the submission.
Learn about the project which made this dataset: https://covid19action-radiology.github.io/index.html
Understand the need for DL models on Radiology for COVID19 Evidence Based Medicine (particularly CXR in Indian context, with challenges on CT): https://youtu.be/l6z5Qo5sVEI