Run the Experiments

1. pip install -r requirements.txt 1.5 Adjust hyper.yaml
2. python3 train.py train.py -> Controls Pytorch Lighnting General Training model.py -> Supervised Model Logic, in Pytorch Lighnting

utils/yaml/XX.yaml configurations utils/utils e.g. loading model, yaml etc. utils/loss e.g. loss loaders utils/augmentations augmentations used

(3.) With a .ckpt model the model can be tested via python3 train.py --mode test

Models Available:

• "resnet18"
• "resnet34"
• "resnet50"
• "vit_small_patch16_224"

Augmentations Available (Code inside ./utils/augmentations.py):

• "no_augmentations"
• "normalization"
• "GeometricAugmentation"
• "ColorAugmentation"
• "ColorAugmentationsChannel"

Losses Available:

• "binary_cross_entropy_with_logits"
• "binary_cross_entropy"
• "cross_entropy"

Changed Torchmetrics.BinaryAccuarcy _safedivide to cpu!!