MSc thesis project at KTH
Using Conda environment, use the requirements.txt file. For EfficientNet, additional steps required described below.
Original code from: https://github.com/lukemelas/EfficientNet-PyTorch.
To install:
cd EfficientNet-Pytorch
pip install -e .
To train example:
python efficientNet.py --epochs 100 --pretrained --batch-size 64 --wd 1e-4 --lr 1.25e-2 --image_size 32 --momentum 0.9 --advprop -val
Arguments and dataset root folder provided in efficientNet.py and file paths defined in parseData.py.
Original code from: https://github.com/toandaominh1997/EfficientDet.Pytorch.
No extra installation required.
Available arguments and file paths provided in each file.
To test that everything is working (EfficientNet pre-trained backbone and EfficientDet):
cd EfficientDet.Pytorch
python test.py
To train example:
cd EfficientDet.Pytorch
python train.py
To evaluate a trained model:
cd EfficientDet.Pytorch
python eval.py
To evaluate and visualize one 2000x2000 WSI:
cd EfficientDet.Pytorch
python evalSlide.py