Submissions will be evaluated based on a multi-factor scoring system that measures both technical performance and practical efficiency. Each model will be scored on the following criteria:

Criteria	Description	Weight (%)
Accuracy / F1 Score	Model’s ability to correctly detect forest fire vs. non-fire instances.	35%
Model Efficiency	Inference speed, memory footprint, and deployability on low-power edge devices.	25%
Novelty / Creativity	Originality of the approach, use of optimization techniques, or architectural innovation.	20%
Reproducibility	Clarity and completeness of code, documentation, and dataset usage description in the notebook.	10%
Environmental Relevance	How effectively the model can be used in real-world forest fire prevention systems.	10%
⚖️ Bonus Considerations

Models that successfully quantize or compress without major accuracy loss will receive extra credit.

Submissions including a working edge deployment demo (e.g., tested on Raspberry Pi, Jetson Nano, or similar) will gain bonus recognition points.

Participants who publicly share their improved dataset version with proper metadata and credits will be highlighted in post-competition results.