geolifeclef2022

GeoLifeCLEF 2022

This repository is related to the GeoLifeCLEF challenge.

The details of the challenge, the data, and all other useful information are present on the challenge page: https://www.kaggle.com/competitions/geolifeclef-2022-lifeclef-2022-fgvc9

Codes

In this repository you will find fully functionnal training/prediction examples on the the GeoLifeCLEF2022 dataset. This includes dataloaders, datamodules, models, training and prediction scripts.

Dataset

To retrieve the dataset, you have two options :

1. Download the dataset manually on Kaggle and decompress the .zip archive in the dataset directory
2. Let Malpolon download the dataset for you via the Kaggle API. To do so, you need to create a Kaggle account, go to your account preferences and create a new Kaggle API token, which is a file named kaggle.json, then place it in your home directory (see here for more details). Finally, you need to change the value of the download_dataset variable in your config file to true.

Training

To run a training script, you simply need to run the following command:

python <name_of_the_script>.py

Several scripts already exist, using different modalities and environmental variables and they use different config files.

• cnn_on_rgb_patches.py and cnn_on_temperature_patches.py use mono_modal_3_channels_model.yaml
• cnn_on_rgb_temperature_patches.py uses homogeneous_multi_modal_model.yaml