U-Net for RGB Semantic Segmentation

Introduction

More details of U-Net architecture can be found on U-Net: Convolutional Networks for Biomedical Image Segmentation. The code was derived from UNet with additional supprt for RGB images.

Dependencies

The following dependencies are needed:

• Numpy
• Tensorflow-gpu
• Keras
• Scikit-image

Usage

1. Install dependencies:

conda env create -f env.yml

2. Configure paths and parameters with config.py
3. Train the model with train.py
4. Test the trained model with test.py

Results