river-segmentation

A Convolutional Neural Network Unet model to segment B&W aerial images into ecological segments.

Installation

1. Clone the project.
2. Install conda (https://docs.conda.io/projects/conda/en/latest/user-guide/install/)
3. Make a new conda enviorment in the terminal: conda create --name myenvname python=3.7
4. Install required packages: conda install --yes --file requirements.txt
5. Activate the enviorment before using the programs in this project: conda activate myenvname

Make prediction with the model

The project includes a trained model that can be used to make predictions on black-and-white areal images of rivers. Download the model (https://drive.google.com/file/d/1GDZ25vxDchf-C_848oPZHNtp7Up7g5NN/view?usp=sharing) To make a prediction: python source/run_predictions.py model_path input_folder output_folder

This project is done for my master thesis at NTNU in Trondheim.