Clear_vs_Cloudy_SimCLR

Binary Classification Model to classify remote sensing imagery as clear or cloudy using the SimCLRv2 approach to learn representations from unlabeled imagery.

LinearEval Notebook

Enables applying the trained SimCLRv2 encoder for a classification task by training a Linear layer attached to the encoder head.

ResNet Baseline

Enables training a supervised ResNet model for a binary classification task.

Installing dependencies

$ conda env create --name simclr --file env.yml
$ conda activate simclr
$ python run.py

Config file

Before running SimCLR, make sure you choose the correct running configurations on the config.yaml file.