Clarification on Model Approach & SAR Integration (Wildfire Project)

[deveshbervar]

Hi @pradeeban ,

I’m applying for the Alaska Wildfire Prediction project and have already contributed PR #66 (Sentinel-2 preprocessing + weather integration + CNN-LSTM skeleton).

While finalizing my proposal, I wanted to ask a higher-level design question:

Given the complexity of integrating multiple data sources (Sentinel-2, SAR, weather, fire labels), would it be preferable to:

1. First establish a strong and well-validated optical + weather baseline model, and then incrementally integrate SAR and additional modalities,
   or
2. Design the pipeline for full multi-source fusion early, even if initial iterations are less stable?

I’m trying to balance research rigor vs engineering stability in early phases.

Any guidance on preferred approach in Alaska projects would be really helpful.

Thanks!
Devesh

[pradeeban]

@deveshbervar I believe you are asking the question in a wrong repository. This is the repository of the Diomede project.

In any case, the deadline is fast approaching. Please make sure to upload your proposal to the GSoC Portal on time.