This repository accompanies our (George Batchkala and Sharib Ali) working-notes paper "Real-time polyp segmentation using U-Net with IoU loss" presented at MediaEval 2020 Multimedia Benchmark Workshop, which was held online on 14-15 December 2020. If you are interested in this work, we recommend you first getting familiar with the overview paper.
To sum up:
Disclaimer: next paragraph was directly copied from the official GitHub repository of the challenge: https://github.com/DebeshJha/2020-MediaEval-Medico-polyp-segmentation.
The “Medico automatic polyp segmentation task” aims to develop computer-aided diagnosis systems for automatic polyp segmentation to detect all types of polyps (for example, irregular polyp, smaller or flat polyps) with high efficiency and accuracy. The main goal of the challenge is to benchmark semantic segmentation algorithms on a publicly available dataset, emphasizing robustness, speed, and generalization.
For more Information consult next section (Information and Links).
train_models.py:
python train_models.py --loss_function="IoULoss" --training_augmentation=0
python train_models.py --loss_function="BCEWithLogitsLoss" --training_augmentation=0
python train_models.py --loss_function="IoUBCELoss" --training_augmentation=0
python train_models.py --loss_function="IoULoss" --training_augmentation=1
python train_models.py --loss_function="BCEWithLogitsLoss" --training_augmentation=1
python train_models.py --loss_function="IoUBCELoss" --training_augmentation=1
checkpoints:
data:
models:
notebooks:
presentation:
submission
train-val-split:
utils:
These are merely for reference.
python=3.8.10
torch=1.10.2 # this is PyTorch (I had it with cudatoolkit 10.2)
torchvision=0.11.3
imageio=2.26.0
torchsummary=1.5.1
torchmetrics=0.9.1