Road segmentation project

Team:

• Natasha Ørregaard Klingenbrunn
• Clara Bonnet
• Daniel-Florin Dosaru

How to train the neural network:
cd scripts
(place the images into training and test_set_images folders)
python3 run.py --train train and predict

Change the value of MODEL_PATH variable in constants.py:
MODEL_PATH = '../best_model.weights'
python3 run.py --predict predict test images

See results by running in the same folder (scripts):
python3 overlayPredictions.py

Install:
For Ubuntu 18.04 LTS run:
sudo apt-get update
sudo apt install python3-pip than ,

Install all the necessary dependencies:

pip3 install -r requirements.txt
sudo apt-get install python3-tk
sudo pip3 install scikit-image

Requirements:

• Python 3.6.9
• torch 1.3.1
• torchvision 0.4.2
• sklearn 0.22
• matplotlib 2.1.0
• scikit-image

Optional

• torchsummary (used to print neural network details)
• psutil (used to print memory usage)