Semantic Segmentation: Unet Trained on Cityscapes Dataset

The Unet model used for image segmentation, ...

Dataset:

The Cityscapes available on https://www.cityscapes-dataset.com/ is a dataset for research purposes only. This dataset contains images from stereo video sequences recorded in street scenes in 50 different cities, as well as the respective annotation of each image.

Requirements:

• python3.8+
• torch
• torchvision
• matplotlib
• opencv
• tqdm

Train the Unet:

python3 main.py

Build and compile the Project:

bash shell.sh
cd build
make

Apply the Model to an image:

./UNet ../Unet-TSM.pt ../test_images/sample-0.png ../RESULTS.png