Mode prediction

Installing packages

Use Conda or Python virtual environment to install python with 3.11 version.

Conda virtual environment

conda create --name <venv_name> 'python==3.11.3'

conda activate <venv_name>

Python virtual environment

python3 -m pip install --user virtualenv

python3.11 -m venv <venv_name>

If python3.11 is not recognized, but python3 points to Python 3.11:

python3 -m venv <venv_name>

source ./<venv_name>/bin/activate  # for Unix/Linux/MacOS

or

.\<venv_name>\Scripts\activate   # for Windows

Install packages

pip install -r requirements.txt

Install dataset

The following script will create an ./input directory in the root of the repository, where it will download the necessary data for training and visualisation.

bash download_all_files.sh 

Train model

The following script will train the model, create a ./experiment directory in the root of the repository where the model weights will be saved. The loss.pdf file with the training plot will be saved in the same directory.

python train_model.py

You can skip train step and download the test model to the ./experiment directory using the following command:

mkdir ./experiment
curl -L $(yadisk-direct https://disk.yandex.ru/d/S5HvchhPkvUd6g) -o ./experiment/model.pth

Visualize prediction

Open visualize.ipynb and run all cells.

Note!

For correct working, directory names must not be changed!