Earthquake Quick Damage Detection using Computer Vision (Turkey-Syria Earthquake Data)

Challenge Background

Turkey is a country between Europe and Asia. It has a large population of 80 million, and many big cities. Last February, the country went through 2 catastrophic earthquakes on the same day (6th of february). During rescue work there were many problems such as detecting the most critical areas and reaching there. The rescue work thus took longer and many lives were put in danger.

The Problem

The reaction time to the latest earthquake was not sufficient, and the rescue teams had to reach many locations in a very short time. Due to the electric shortage in the area, communication was problematic for damage assessment. Thus, a fast-responding AI model is planned to be used to aid rescue operations planning. This model will detect earthquake damage according to the latest satellite images provided by online service providers. The damage search will include building damage, road damage, and terrain changes. Also, these changed areas on the map will mark the changes’ locations and classify these changes as building, road, or terrain. The model is planned to use mainly Turkey-Syria Earthquake data.

Goal of the Project

• Develop a model with functions to find the changed locations after a catastrophic earthquake.
• Employ cutting-edge technologies like computer vision and deep learning techniques to improve the speed and accuracy of detecting damage and responding.
• Give the model functionality to classify different types of damages such as building, road, terrain change, etc.
• (Optional) Develop an API for the model.

Project Timeline

What you'll learn

1. Extract and collect remote sensed data from Google Earth Engine
2. Analyze remote sensed data to reveal insights and identify trends.
3. Perform image processing and segmentation to extract earthquake-affected areas.
4. Implement computer vision models to classify the damage type.