Inspiration
Every major conflict requires time-sensitive decisions, whether of natural causes (floods, wildfires, earthquakes) or human (warzone, terrorist attacks, major malfunctioning). The recent aerial strikes on densely populated areas in Ukraine prompted a necessary reminder of how well-coordinated emergency responses can significantly reduce casualty, inefficiencies, and potentially save lives in a matter of minutes.
What it does
A well-formed coordinated response during public crises involves having mastery over two factors: data and communication. Leveraging data: OpenResponse maps the city of Mariupol, Ukraine in terms of how damaged its buildings are as of March 21, 2022, at the beginning of the Russian-Ukraine conflict. A city of approximately 430 000, we make it clear which areas are in heavy distress, which are not, and visualize the physical impact from a bird's eye view all from the browser. Enabling communication: On top of a rendered map, the streets of Mariupol are entirely navigable. From any starting point, users can see optimal routing to virtually any area in a single click.
How we built it
Working with Geolocation We used a robust location data platform called Mapbox to create the customized mapping & navigation environment. Both the design & the backend are written in JavaScript and React with HTML/CSS.
Working with Satellite Imagery Data We took apart a recent "dataset identifying damage sustained in Mariupol as of March 21st, 2022" by Scale AI. What was initially 3481 satellite images with the boundary of each roof labeled with a polygon, we parsed the data to obtain over 81 000 structures each with its own metadata of damage assessment and coordinates after several coordinate system conversions. Using geoJSON, we imported all structures into Mapbox for labeling.
Designing the Interface We simplified the interface to highlight critical damage identification. All structures of Mariupol are either labeled "Undamaged", "Damaged", or "Destroyed" with the appropriate color. Paramount infrastructures such as hospitals and schools are specially color-coded.
Challenges we ran into
Oh My Dataset! Scale AI's Mariupol Damage Assessment dataset was released at the onset of the Russian-Ukrainian conflict to the immediacy of its potential application. An obstacle later revealed was the lack of documentation for non-industry developers working with geolocation data and mapping. Many coordinate system conversions are left implicit such as that from the traditional Latitude/Longitude system you'll find on Google Maps to the less-intuitive Universal Transverse Mercator (UTM) coordinate system or the World Geodetic System 1984 (WGS84). It was not before numerous realizations while parsing the dataset that we came to see just how little we knew about the art of mapping n' modeling the geographical complexity of the ellipsoid we so call Earth.
What we learned
Working with Geodata was undoubtedly a feat, during both the parsing stage and the mapping & marking stage. Who knew putting coordinates on a map was more than... putting coordinates on a map? Conventional coordinate systems, conversion, calibration, and polygon labeling, were completely novel concepts to tackle. For some of us, working with React was also a new environment.
Some Over Others An important lesson from a non-technical perspective was just how much was left off the table in the making of this prototype. Datasets as such provide many generalized ideas for what to do from a technical POV, but focusing on a few core features such as making it interactive, navigable, and eventually open for collaboration for the right users in mind (not us developers) can be overlooked at times.
What's next for OpenResponse
We plan on deploying an MVP (web app) incorporating shorter refresh cycles of satellite imaging and much larger datasets of damage assessment and resource distribution during crises, as well as authenticated user types (medics, civilian rescue, journalists, etc). This would involve extensive cooperation with logistics experts, emergency response personnel, and rescue operations coordinators to identify and validate key problems that can be solved on a local scale with measurable human and financial impact.
Looking Ahead & Anticipate
Scaling Actors & Coordination What is absent in our current version but is very feasible for an upcoming version is the addition of user groups with live geolocation. It's useful to think of products like Snapmap where users can opt-in on broadcasting their location. OpenResponse can then house live locations & routes for Emergency Medical Response teams, firefighters, Civilian Rescue squads, or even journalists who can coordinate their efforts seamlessly. Additional layers of non-location user groups such as logistics coordinators, local and state emergency operations directors, etc. The more agencies are involved, the better the response, rescue, and rebuilding.
Scaling Coverage OpenResponse's data and mapping tools can easily scale to serve conflicted regions globally. The novel capabilities of satellite imagery combined with machine-human labeled datasets from the likes of Scale increase the coverage for potential disaster relief in rural or disconnected populations, proving its resilience as a tool for gathering damage assessment and geolocation data for those who need it the most.
Scaling Data & Prevention As more data on various types of conflicts are gathered to build predictive models, it becomes possible to predict flood patterns, direction of wildfire spread, explosion patterns, optimal rescue routes, efficient resource distribution, and many more capabilities we've yet to even fathom.
Log in or sign up for Devpost to join the conversation.