Kneightbros.AI| Hey! What Are You Looking At?
Hey! What Are You Looking At?
Cyclon.AI
Summary
Our project seeks to develop a new and innovative hurricane detection methodology based on deep learning. Our algorithm is trained with images provided by NASA's aqua and terra satellites.
How We Addressed This Challenge
What did you develop?
Hurricanes currently pose an imminent threat of natural disaster. The early detection of hurricanes through deep learning will allow the implementation of preventive measures with more anticipation and with valuable information to determine the impact characteristics of the hurricane. In addition, this system can be automated to develop an alert system for predictions.
Why is it important?
The detection of hurricanes provides a series of deep-rooted benefits to the prevention and mitigation of damages that these weather events generate. Also, artificial intelligence models like this one can be easily scaled to detect other events.
What does it do?
Cyclone.AI is an artificial intelligence application capable of receiving satellite images of rainfall on the globe on a friendly platform that allows detection and predictions of possible hurricanes on earth.
How does it work?
It is an artificial intelligence that took images of the "GIBS API" of NASA's Terra & Aqua missions. The photographic images were taken from the year 2000 to the present year taking into account all the most known hurricanes to date and recorded in the images to build the training model in a convolutional neural network of ResNet architecture.
What do you hope to achieve?
The scope of our project that is currently a powerful prototype and scalable enough to integrate on different platforms. Our ideal is to integrate it directly into the decision making of geospatial earth monitoring platforms and receive video analysis, images in real time. Eventually, correlation models will be inserted that can indicate the economic impact with the size of the hurricane detected, the trajectory prediction, and variables of the country to which it is heading.
How We Developed This Project
What inspired your team to choose this challenge?
Initially, we value taking advantage of the information generated by NASA's satellites to mitigate losses due to natural disasters and be prepared, taking into account the damage caused when a Hurricane is inserted on land, economic expenses can be reduced and better prepared for contingencies mainly in an automated way not only for NASA and geospatial agencies but also for humanity through different platforms to which they have access.
What was your approach to developing this project?
Predict and detect the formation of hurricanes in an automated way through artificial intelligence with images taken from space missions on precipitation on Earth.
What tools, coding languages, hardware, software did you use to develop your project?
We used an electronic whiteboard for the problem understanding phase to brainstorm about the possibilities we have with the information provided by NASA and GBIS. For the structuring and evaluation of the project for 8 hours we use the agile Kanban methodology. The programming language with which the project is written is Python through the Tensorflow API to extract architecture from artificial intelligence models for the construction of the prototype and validate its feasibility. The hardware used was an Nvidia Geforce GTX1050TI graphics card to speed up the training process through Pycharm's programming and development interface and streamline framework for web demonstration.
What problems and achievements did your team have?
At the team level, communication and creativity are somewhat affected by social distancing, but it was a phase that we managed to overcome, however, the biggest problem was the extraction and conversion of data from the GIBS satellite images to use. In our analysis, it took more than 70\% of our time, but having well-structured data the results are extraordinary and do not represent an obstacle to the development of the project.
How We Used Space Agency Data in This Project
The data played a fundamental role in the development of the artificial intelligence model and its subsequent training, thanks to the images (they were extracted from GIBS, NASA's Terra and Aqua missions). The process carried the conversion of the format used for the visualization in the GIBS platform to jpg, for easy use of images in the model that were segmented to recognize the formation of hurricanes from the images.
Project Demo
Google Drive Video:
https://drive.google.com/file/d/1dc2bw2rcwqk3q7f06kKM07wK5SnHJ6DI/view?usp=sharing
Youtube Video
Project Code
Data & Resources
Global Imagery Browse Services (GIBS). 2000-2020
Copyright © 2013 - 2020 United States Government as represented by the Administrator of the National Aeronautics and Space Administration. All Rights Reserved. This software is licensed under the NASA Open Source Software Agreement, Version 1.3. Source code is available on the NASA GIBS GitHub.