TensorForest 🌲🔥

Innovating Wildfire Prevention with Drones, LiDAR, and Machine Learning

TensorForest proudly secured the top position at the WRO (World Robot Olympiad) Toronto Regional Competition in the Future Innovators category, advancing to the national level. Additionally, the project received a $5000 grant from the Youth Climate Action Fund, run by the Town of Oakville in collaboration with the Bloomberg Foundation.

TensorForest is a cutting-edge Python-based application designed to tackle the escalating threat of wildfires. Developed by Arnnav Kudale, Shayaan Azeem, and Rohanth Marem, TensorForest combines automated drone technology, LiDAR, and machine learning to provide a robust, efficient solution for wildfire prevention and monitoring. This project reflects our team's passion for merging robotics with environmental sustainability.

Wildfires are a growing environmental concern, causing devastating damage to ecosystems, wildlife, and human communities. Early detection and precise monitoring are critical to mitigating their impact. Existing methods often fall short in accessibility, affordability, or accuracy. TensorForest aims to bridge this gap by offering an innovative solution tailored to real-world conditions.

• LiDAR Integration: TensorForest now includes LiDAR technology to map tree canopy heights, enabling more detailed vegetation analysis.
• Flammability Mapping: Custom machine learning models assess vegetation types and flammability levels to predict wildfire risks.
• NDVI Data Mapping: Utilizing a GoPro 7 with a Backbone mod to capture full-spectrum imagery and generate NDVI (Normalized Difference Vegetation Index) data for vegetation health analysis.
• Real-Time Insights: Drone-collected data is processed to generate actionable wildfire prevention strategies.

1. Hardware Integration: Ensuring seamless compatibility between flight controllers, LiDAR sensors, GPS modules, and camera systems.
2. Software Complexity: Developing machine learning models, flight planning algorithms, and image processing scripts.
3. Field Testing: Managing logistical and safety challenges during real-world drone testing in forest environments.

• Drone Hardware: Built using a lightweight, prebuilt carbon fiber kit with custom components for optimal performance.
• LiDAR Technology: Provides precise tree canopy height mapping for detailed vegetation analysis.
• Full-Spectrum Imaging: Captured with a modified GoPro 7 to generate NDVI data for vegetation health mapping.
• Machine Learning: TensorFlow models trained to classify vegetation and analyze wildfire risk factors.
• Open-Source Software: Leveraging ArduPilot for autonomous drone navigation and TensorFlow Lite for object detection.

1. Topography and Heat Maps: Incorporating topographic mapping with real-time heat mapping for comprehensive wildfire risk analysis.
2. Model Redesign: Refining machine learning algorithms for enhanced accuracy and scalability.
3. Community Collaboration: Partnering with authorities and environmental organizations to deploy TensorForest in wildfire-prone areas.

• Arnnav Kudale – @blazecoding2009
• Shayaan Azeem – @shayaan_azeem
• Rohanth Marem – @rohanthmarem

Project Link: TensorForest on GitHub

TensorForest was made possible thanks to the support of the Youth Climate Action Fund, Mayor Rob Burton, and the guidance of mentors, educators, and community members who believed in our mission to revolutionize wildfire prevention.