Beat-the-MAML Game

This app showcases a gamified survey interface that we used in our paper "Humans are poor few-shot classifiers for Sentinel-2 land cover", presented at the IGARSS 2022 conference.

Play with the app at beat-the-maml.westeurope.cloudapp.azure.com

---

please consider citing

@inproceedings{humanfewshot,
  title={Humans are poor few-shot classifiers for Sentinel-2 land cover},
  author={Ru{\ss}wurm, Marc and Wang, Sherrie and Tuia, Devis},
  booktitle={Proceedings of 2022 IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2022)},
  year={2022}
}

alongside prior work on "Meta-Learning for Few-Shot Land Cover Classification"

@inproceedings{russwurm2020meta,
  title={Meta-learning for few-shot land cover classification},
  author={Ru{\ss}wurm, Marc and Wang, Sherrie and Korner, Marco and Lobell, David},
  booktitle={Proceedings of the ieee/cvf conference on computer vision and pattern recognition workshops},
  pages={200--201},
  year={2020}
}

if you derive scintific work from our studies.

Run

The run infrastructure is based on the EPFL-ENAC helloFlask template to run:

• locally with Flask command (dev)
• locally with Gunicorn (test)
• locally or on server with Gunicorn inside Docker (prod)

locally

Run the following commands :

make setup # only once
make generate-selfsigned-cert # only once

make run-dev # dev mode (Flask)
make run-prod # test mode (Gunicorn)
make run # prod mode (Gunicorn inside Docker)

on server

Run the following command :

make run # prod mode (Gunicorn inside Docker)

Stop the service :

docker-compose down

Check result

• dev / test : browse to http://localhost:5000
• prod : browse to https://localhost (you'll have to accept self signed certificate)

License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.