Chariot| Home Planet at Your Fingertips
Home Planet at Your Fingertips
Maria System: Analyze.Explore.Visualize
Summary
Maria System is an application with which any user can analyze data from space agencies or user's own datesets, and each user has the ability to visualize them via graphs, maps, etc. User can make predictions based on polynomial regressions, logistic regressions, cross-validations, k-mean clustering, KNN, r^2 data comparison, etc.The app has 5 languages: English, Hindi, Russian,Arabic and Chinese. Our innovative module for SDK developed by us allow to add new language in few clicks. Translation is based on machine learning.Developers and researchers can use our API to add our data analyzing functions (especially ML) to their own projects.
How We Addressed This Challenge
Maria System is a special application and API (we named it Maria System EasyML) for developers and researchers that allows you to analyze, visualize and solve such types of problems as: polynomial regressions, K-nearest neighbors (KNN), k-means clustering, series comparison, image recognition.
Polynomial regressions performs the task to predict a dependent variable value (y) based on a given independent variable (x) based on previous x and y arrays. It can be used to predict next values in timeseries dataset. For example, Carbon Dioxide in Amazon Forests time series, global warming, etc.
K-nearest neighbors (KNN) assumes the similarity between the new case / data and available cases and put the new case into the category that is most similar to the available categories. For example, determine the types of new objects, new animals categorization, earthquake predictions, etc.
K-means clustering Identifies k number of centroids, and then allocates every data point to the nearest cluster, while keeping the centroids as small as possible . For example, split the uncategorized data by different clusters. It can be used to fine anomalies in data and make analysis of absolutely new uncategorized (unlabeled) data.
Series Comparison allow easily compare two or more series for similarity. For example, you can use it to find the correlation between different data like how much global warming data correlate with CO2 reduce.
Image recognition allows you user offline!!! load their own trained models to our mobile app and used it, for example, to recognize object on the satellite images.
Researches can analyzing their data on our server thought our API.
Data processing and analysis is challenging for many people, including scientists, engineers, and developers, especially if they have low-performance computer. We hope that our solution will help researchers and developers in in exploring our planet, the main task of the project is aimed at ensuring that people of these professions do not spend a lot of time processing data.
How We Developed This Project
We chosed "Home Planet at Your Fingertips" this challenge because we believe that data processing anywhere in the world is an extremely necessary thing for the researchers of our planet. With the help of data processing, it is possible to find out where and when a natural disaster, fire, flood, etc.
Communaction:
ZOOM, GitHub;
Design:
Figma, Adobe PhotoShop;
Server side:
Programming languages: Python 3.8, C++;
OS: Ubuntu 18.04;
Hosting: Digital Ocean UK droplet;
NGINX server, FastAPI, Flask;
Mysql DB;
Let's encrypt for SSL certificate for secured https connection;
AES256 encryption for data on server protection;
FileZilla for sftp connection for data transferring to the server.
Pentest-tools.com to check potential vulnerabilities of our solution.
Client side:
Anaconda python SDK;
Xcode sdk;
Programming languages: Python3.8, Swift, Object-C;
danielgindi charts designed views swift
Apple TestFlight;
Hardware:
Raspberry Pi 3 model B+
What we achieved during two days of the hackathon:
We developed a working mobile app for IOS devices, working API for developers and researchers. After we developed our app, we have small research on Google Patent Base, Scholarship and PatBase, and we found that some of our solutions is unique. For example, translation app to different languages in a few clicks, algorithm to compares combinations in pairs between them, which can significantly speed up some of the data processing tasks. We also have research on different stores and we didn't find any app, which allows all of our functions.
Problems:
We didn't have enough time to finish Android App, add AR visualization.
We sent our mobile app to AppStore review, but in the October 4, the status of the app is still "Waiting for review". However, users can use and test our mobile app via TestFlight.
How We Used Space Agency Data in This Project
Included in the app:
NASA Datasets: Meteorits landing - clustering and map visualization
Nasa Earth data: Satellite images - image recognition
Nasa Earth data: Global Fire Emissions Database, Version 4.1 (GFEDv4). - clustering and map visualization. Also, researchers can used KNN to categorize new fairs.
Nasa Earth data: LBA-ECO CD-02 Carbon and Oxygen Isotopes in Atmospheric CO2 in the Amazon: 1999-2004.
Nasa Earth data: NACP Peatland Landcover Type and Wildfire Burn Severity Maps, Alberta, Canada.
Special demo for API:
Nasa Earth data with modification from data from World Data bank: Global Warming data, global CO2 reduce data during over 30 years. Special demo for data comparison and finding the score of correlation.
Also, our API allows to user analyze another datasets from NASA's partner agencies.
Project Demo
Presentation: https://drive.google.com/file/d/16_7bH1kpC1we1x5c0yoL-EnbRBaQxtm1/view?usp=sharing
Demo of mobile app (because of 30 second limit, we didn't have enough time to show all of our mobile app functions): https://drive.google.com/file/d/1Ibuw_EB5YQCp44QayKh6Qsv8Fkw2std2/view?usp=sharing
Also you can fill the form and we send the invite to the TestFlight of our app: https://docs.google.com/forms/d/e/1FAIpQLSfCy9QLUG_-Rahj32ssoCFxZ7ANqXatb4ySDjDosrhPembZRw/viewform
Demo of API (Maria System EasyML) you can see in our GitHub repository: https://github.com/Chariotteam/MariaSystem
Project Code
Data & Resources
NASA Datasets (data.nasa.gov);
NASA EarthData;
Kaggle;
World Data Bank;
MIT research;
Google Patent Search
PatBase