Climate Hack.AI (2022) Submission Validator
This repository contains code that allows you to quickly validate your Climate Hack.AI (2022) submission locally.
targets.npz contain EUMETSAT HRV Satellite Imagery data preprocessed by Open Climate Fix and further processed into the format used for evaluation in Climate Hack.AI 2022, an international datathon between the student communities of 25 world-leading universities.
This validator uses 200 image sequences taken between 10am and 4pm over Great Britain from days sampled uniformly at random between January 2020 and November 2021 to score your submission.
Instructions for use
Ensure you have NumPy, PyTorch and
pytorch_msssiminstalled (as well as the other dependencies listed in the Climate Hack.AI getting started project).
validate.pyfile to your existing submission folder, and the
targets.npzfiles one level up.
Change into the
validate.pyusing Python 3.9 (e.g.
Note: this assumes the script specified in your
doxa.yaml file is
evaluate.py as by default, so you will need to modify
validate.py if you have changed it.