Awards & Nominations
Saudi Space Shuttle has received the following awards and nominations. Way to go!
Global Nominee
ML-Based Marine Spill Detector
High-Level Project Summary
We met and exchanged our diverse science background. We were able to collect over a thousand short clips of the marine chemical and oil spills. The samples are basically videos that span one month before the incidents to one month after the incidents, so the machine learning model can detect the pollution emergency type. The model is trained on imagery layers of satellites radar readings. The metadata of the readings had been obtained from the NOAA historical data, and our approach is exposed to oil spill data since 2014 till now, since SAR is highly utilized due to its detection abilities. Therefore, our web service can scan the globe and generate updates on maybe a daily basis.
Link to Project "Demo"
Link to Final Project
Detailed Project Description
We propose an oil spill detection web application that periodically reads ESA and NASA satellite data and generates a global geographical heatmap. The purpose of the heatmap is to indicate the occurrence chance of the spillage incidents using regular deep learning methods. Our method mainly consists of combining open source datasets for the NASA Space Apps to help natural disasters emergency responders and decision makers plan and allocate resources efficiently. Our vision is to have the contamination accumulation in the food chain de-accelerated, and the mission of the public user interface (SaudiSpaceShuttle.com) is to provide the the first responders community with an additional deep model as well as to increase the public awareness about water pollution.
Space Agency Data
In the ML-Based Marine Spill Detector, several pieces of open data are utilized, and two datasets are joined for extracting the main features for the marine oil-spill, chemical, and other incidents. The regular satellite imagery of the Earth API was prone to cloud blocking. Accordingly, the Synthetic Aperture Radar (SAR), from European Space Agency (ESA), which has a long-lasting partnership with NASA, was practical since it is less variant to weather conditions, and so it has multiple applications, one of which is oil pollution detection. In 2023, a new SAR addition is expected to be added by NASA's joint NASA ISRO SAR Mission (NISAR) program. In any case, the ability of reading the satellite imagery and radar readings isn't adequate without the historical data of the marine pollution incidents. For that reason, the ESA dataset was joined with the Raw Incident Data, which is open and is accessible via the IncidentNews website of National Oceanic and Atmospheric Administration (NOAA). In 2016. NASA launched an advanced weather satellite for NOAA. Nevertheless, the ESA data, along with several other datasets, was accessible through Google's Earth Engine API. It needed filtering libraries, available in Python and JavaScript, to prepare the data for downloads, and so does the Earth API. We joined the NOAA dataset to ESA data to produce the imagery data, which becomes open to the public when it is published, for our deep learning model. Our model also trains on the pre-processed ROBORDER data from MultiMoDal Data Fusion and Analytics Group. The ROBORDER data is accessible upon requesting the research group. Please find the list of the utilized data below.
The utilized data:
- NASA, Earth API, [URL] https://api.nasa.gov
- ESA, Google Earth Engine API, [URL] https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S1_GRD
- NOAA, Raw Incident Data, [URL] https://incidentnews.noaa.gov/raw/index
- "MultiMoDal Data Fusion and Analytics Group," ROBORDER Dataset, [URL] https://m4d.iti.gr/oil-spill-detection-dataset
Hackathon Journey
The journey was exciting but too short. The NASA Space Apps had many interesting data, and it will have to be revisited. Thank God, we recruited excellent talents, and our team was neat. We met inspiring team members, discussing the approaches. However, the challenge reminded us that life is too short, like the NASA Space Apps challenge. We also contacted other researchers, who shared their newly collected dataset. We utilized Google Compute Engine Platform to accelerate the work. Our team had amazing communication skills, and we had a great time cooperation.
The challenge was well organized, and we were helped throughout the journey.
Our introductory video was called ML-Based Marine Spill Detector (1), and it is at https://youtu.be/M0WUgLofMjE.
Our second video is here https://youtu.be/2gkBq08VYaQ.
Our Github page: https://github.com/Saudi-Space-Shuttle/ML_Based_Marine_Spill_Detector
Our workplace folder: [available upon request]
The first step was to collect and acquire the data from several places. The second step was cleaning the data to make sure its pieces are consistent. In the third step, we utilized the satellite radar imagery to extract the features. Then we started the deep learning phase, where we are fine tuning the hyper parameters. We didn't finish this step yet, but our team has prepared the basic elements of the graphical user interface.
References
1. Krestenitis, M., Orfanidis, G., Ioannidis, K., Avgerinakis, K., Vrochidis, S., & Kompatsiaris, I. (2019). Oil Spill Identification from Satellite Images Using Deep Neural Networks. Remote Sensing, 11(15), 1762. 2. Krestenitis, M., Orfanidis, G., Ioannidis, K., Avgerinakis, K., Vrochidis, S., & Kompatsiaris, I. (2019, January). Early Identification of Oil Spills in Satellite Images Using Deep CNNs. In International Conference on Multimedia Modeling (pp. 424-435). Springer, Cham.
Tags
#Collaboration
Global Judging
This project has been submitted for consideration during the Judging process.