Road Segmentation for RoboTour 2024

This repository contains the code for training and testing a road segmentation model for the RoboTour 2024 competition.

Installation and setup

The code is meant to be developed locally in a Python 3.10 environment and then run on a remote server. Before we delve into the local and remote setup, we have to sign up to the Weights & Biases platform to log the training and testing process. After signing up, log in to the platform via the terminal by running the following command:

wandb login

This process is described in the official documentation.

Local setup

1. Clone the repository and navigate to the root directory of the project.

   git clone https://github.com/aleskucera/road-segmentation.git
   cd road-segmentation

2. Install PyTorch by following the instructions on the official website.
3. Install other requirements by running the following command:

   pip install -r requirements.txt

Remote setup

Remote machines should use Slurm Workload Manager to manage jobs. The following steps are for setting up the remote

1. Clone the repository and navigate to the root directory of the project.

   git clone https://github.com/aleskucera/road-segmentation
   cd road-segmentation

Note

There is no need to install any requirements on the remote machine as the dependencies are loaded in the Slurm script. The script will load the modules required for the job. Check the .env file for the list of modules.

Usage

Local usage

The code is divided into two main parts: training and testing. To run training on the local machine, run the following command:

python main.py action=train

To run testing on the local machine, run the following command:

python main.py action=test ckpt_path=[checkpoint path]

You can also override the default configuration like dataset by passing the desired configuration as a command line argument. For example, to change the dataset to RoboTour, run the following command:

python main.py ds=robotour

For more details about default configuration and override options, run the following command:

python main.py --help

Remote usage

To run training on the remote machine, run the following command:

sbatch [sbatch options] scripts/train.batch [script options]

To run testing on the remote machine, run the following command:

 sbatch [sbatch options] scripts/test.batch [script options]

For more details on the options, run the following command:

bash scripts/train.batch --help

or

bash scripts/test.batch --help