Discussion: Alaska Wildfire Prediction Using Satellite Imagery (GSoC 2026) #68

Hi everyone,

I’m Eshaan Saha, a B.Tech student with a strong interest in machine learning, computer vision, and climate-focused AI applications.

I’m particularly interested in the "Alaska Wildfire Prediction Using Satellite Imagery" project. I have prior experience working with spatiotemporal data through an EEG signal analysis project during my internship at IIT Jodhpur, as well as a computer vision project involving UAV-based fruit ripeness detection using YOLOv8 with GPS mapping.

I’m currently exploring approaches for wildfire prediction using multi-temporal satellite imagery combined with environmental data (e.g., temperature, humidity, wind). I’m considering a CNN + LSTM/ConvLSTM-based architecture for spatio-temporal modeling.

I would love to get your thoughts on:

• Preferred datasets (MODIS, VIIRS, Sentinel?)
• Whether temporal prediction is expected beyond detection
• Any existing baselines or prior work in this repo

Looking forward to contributing and learning from the community!

Thanks!