Issue 1

[weiji14]

Background

Change detection for Burned area Delineation (ChaBuD) ECML/PKDD 2023 challenge

• Main website: https://2023.ecmlpkdd.org/submissions/discovery-challenge/challenges
• HuggingFace Spaces: https://2023.ecmlpkdd.org/submissions/discovery-challenge/challenges/
• Datasets:
  • https://huggingface.co/datasets/chabud-team/chabud-ecml-pkdd2023/tree/main
  • https://huggingface.co/datasets/chabud-team/chabud-extra/tree/main

Goals

• Open source repo showcasing how to perform change detection of burned areas from optical imagery
• Get a blog post (or series of blog posts) out on the data pipeline, machine learning model and results

Partner and Stakeholders

Development undertaken by Gaia pod @developmentseed, but open to contributions from other interested parties!

Team

• OD @weiji14
• D @srmsoumya
• D ???

Milestones

Timeline from https://huggingface.co/spaces/competitions/ChaBuD-ECML-PKDD2023:

• April 11, 2023: competition starts. Train and validation sets are released
• May 8, 2023: Submission opens! Public leaderboard is made available
• June 9, 2023: hidden test set is released (input data only). Private leaderboard is made available.
• June 18, 2023: end of the challenge
• June 25, 2023: final leaderboard is released
• July 15, 2023: deadline for final report submission
• July 31, 2023: final decision and winners are announced

Project Organization

• GitHub Repo: https://github.com/developmentseed/chabud2023
• HuggingFace Spaces: ???

Documents

• Registration at https://forms.gle/oFXcJHnNaVKnvKak8

[weiji14]

Resources

• TinyCD
  • Codegoni, A., Lombardi, G., & Ferrari, A. (2022). TINYCD: A (Not So) Deep Learning Model For Change Detection (arXiv:2207.13159). arXiv. https://doi.org/10.48550/arXiv.2207.13159
  • Code: https://github.com/AndreaCodegoni/Tiny_model_4_CD
• DS-UNet
  • Monaco, S., Greco, S., Farasin, A., Colomba, L., Apiletti, D., Garza, P., Cerquitelli, T., & Baralis, E. (2021). Attention to Fires: Multi-Channel Deep Learning Models for Wildfire Severity Prediction. Applied Sciences, 11(22), 11060. https://doi.org/10.3390/app112211060
  • Code: https://github.com/dbdmg/rescue