A segmentation model has been developed with ability to use multiple loss function options and customizable arguments. The model supports several configurations, including a flat U-Net, as well as U-Net variants with ResNet-34 and ResNet-50 encoders.
To ensure compatibility across different environments, the entire project has been containerized using Docker. This allows for a plug-and-play approach, simplifying the process of running the model in various setups.
Will be added later
Will be added later
git clone https://github.com/Akaqox/unet-segmentation-with-docker.git
cd unet-segmentation-with-docker
docker build -t seg:latest .
docker run -v /opt/data/seg:/app/results --gpus all -it --ipc=host seg
python -u main.py --bs --model unet50
python -u inference
python -u inference --image 'path to image'
python -u inference --jv