🏆 EDGEFIRE CHALLENGE 2025 — OFFICIAL COMPETITION RULES
ENTRY IN THIS COMPETITION CONSTITUTES YOUR ACCEPTANCE OF THESE OFFICIAL COMPETITION RULES.

See Section 3.18 for defined terms

The EdgeFire Challenge 2025 is a skills-based competition designed to promote innovation in Edge AI and Environmental Monitoring. You must register via the Competition Website to enter. To participate, you must agree to these Official Competition Rules, which include all terms, content, and policies referenced on the Competition Website and within these Rules (collectively, the “Rules”). By submitting an entry, you agree to be bound by these Rules. You may not submit to the Competition and are not eligible for any prizes unless you agree to these Rules. These Rules form a binding legal agreement between you and the Competition Sponsor.

1. COMPETITION-SPECIFIC TERMS
1. Competition Title

EdgeFire Challenge 2025 — Edge-Friendly Forest Fire Detection

2. Competition Sponsor

National Center of Artificial Intelligence (NCAI), NUST University
in collaboration with ConservAi Research Initiative

3. Competition Sponsor Address

National Center of Artificial Intelligence (NCAI),
NUST Campus, H-12, Islamabad, Pakistan

4. Competition Website

https://www.kaggle.com/competitions/edgefire-challenge-2025

5. Total Prizes Available:

$600 (Equivalent in PKR)

🥇 First Prize: $300

🥈 Second Prize: $200

🥉 Third Prize: $100
(Non-monetary recognition certificates and publication opportunities for Top 10 teams)

6. Winner License Type

Open Source License (MIT License)

7. Data Access and Use

Competition Use and Non-Commercial & Academic Research
Participants may access and use the dataset for non-commercial, educational, and research purposes, as described below.

2. COMPETITION-SPECIFIC RULES
1. TEAM LIMITS

a. Maximum team size is five (5) members.
b. Team mergers are allowed before the Team Merger Deadline, provided the merged team’s total submission count remains within the competition limit.

2. SUBMISSION LIMITS

a. Each team may submit up to five (5) submissions per day.
b. Teams may select up to two (2) final submissions for evaluation.

3. COMPETITION TIMELINE

All official competition dates (Start Date, Entry Deadline, Final Submission Deadline, and Team Merger Deadline) are available on the Timeline page of the competition website.

4. COMPETITION DATA
a. Data Access and Use

The dataset provided for this competition is the Forest Fire Dataset for Binary Classification, consisting of labeled fire and nofire image data.
Participants are permitted to:

Download and use the dataset for model development and submission.

Shuffle, preprocess, or augment the data to improve model performance.

Create and publish modified public versions of the dataset for research purposes, provided proper credit is given to the original dataset and changes are documented.

Use the dataset solely for non-commercial, educational, and academic research purposes.

Note: Redistribution or sale of the dataset for commercial purposes is strictly prohibited.

b. Data Security

Participants must:

Ensure that only authorized team members have access to the dataset.

Prevent unauthorized duplication or sharing of the dataset.

Notify the organizers immediately of any data misuse or breach.

5. WINNER LICENSE

All winning submissions will be released under the MIT Open Source License, allowing open use for academic and research purposes.
By submitting, you agree that:

Your source code and model architecture may be publicly shared for educational transparency.

The organizers may use or demonstrate your model (with attribution) in environmental or academic applications.

6. EXTERNAL DATA AND TOOLS

Participants may use external datasets and open-source pretrained models under these conditions:

The external data is publicly available, free to access, and equally accessible to all participants.

Any external data or pretrained models used must be clearly documented in the submission notebook.

The use of large proprietary or paid datasets is not allowed unless approved by the organizers.

Allowed Tools:

TensorFlow, PyTorch, Keras, ONNX, Scikit-learn

Optimization libraries for pruning, quantization, and model compression

Edge deployment libraries such as TensorFlow Lite, OpenVINO, or PyTorch Mobile

Automated Machine Learning Tools (AutoML) may be used if the participant holds valid licenses and documents the workflow.

7. ELIGIBILITY

Open to students, beginners, and AI enthusiasts worldwide aged 16 or older.

Employees, interns, or contractors of the Competition Sponsor or affiliated institutions may participate but are not eligible for prizes.

8. WINNER’S OBLIGATIONS

Winners must:

Submit their final model code, documentation, and preprocessing pipeline.

Provide a reproducible notebook showing all data handling, training, and evaluation steps.

License the winning solution under the MIT License.

Complete all required forms (eligibility certification, tax forms if applicable).

Failure to comply may result in disqualification.