Deep Hazard| Automated Detection of Hazards
Awards & Nominations
Deep Hazard has received the following awards and nominations. Way to go!
Automated Detection of Hazards
Deep Hazard - generate hazard maps automatically by deep learning
Summary
We have created a system that automatically generates hazard maps, only by geographical characteristics from google earth engine. It could be applied to areas where hazard maps are not prepared enough. In such areas, people including foreigners have difficulty in acknowledging the risks they can take.Hazard maps can encourage local people to be more aware of natural dangers. Furthermore, foreigners will be able to travel and settle in that country more at ease. Our goal is to make the world more secure to live in. In order to achieve this, we plan to roll our system out on a global scale as an application that is accessible from all over the world.
How We Addressed This Challenge
We have developed a machine learning model that can predict the potential risks of damage caused by natural disasters. In short, it generateshazard maps automatically.
This system only needs geographical information such as height, vegetation obtained from google earth engine.
It is very important to have reliable hazard maps because it will raise awareness of disaster prevention among people living there, which in turn will help governments to make decisions on disaster prevention policies and reduce the damage caused by disasters.
However, there are many areas in the world where hazard maps are not well organized. If available, they are often not based on appropriate standards or are not easily accessible to foreigners.
Therefore, we have developed a system to improve such situations.
For example, even in developing countries where information about disasters is scarce,
our system makes it possible to obtain hazard maps for the entire region, by using data from geographically similar areas.
Hazard maps are helpful not only for the local but for foreigners, because they will be able to travel or settle in that country more at ease.
Moreover, when humans move to an unknown planet in the future, we may be able to predict the disaster risk for the entire planet based on several years of preliminary research in the limited area.
If this is realized, our system will undoubtedly make a significant contribution to the humanity's future.
Our system automatically generates hazard maps from geographic data in the google earth engine.
The supervised data used for training of the model are the hazard maps in Japan provided by the Geospatial Information Authority of Japan. They include information about floods, tsunamis, landslides.
We conducted a small-scale training and test on the data in Japan and confirmed that the training was possible with a certain degree of accuracy. The data we used include elevation, vegetation, land cover, river flow, vegetation and slope.
Our goal is to make the world more secure to live in. In order to achieve this, we plan to roll our system out on a global scale as an application that is accessible from all over the world.
How We Developed This Project
Our team consists of Japanese, and in Japan, a lot of natural disasters take place such as earthquakes, floods, tsunamis, and so on. Therefore when we look for a place to live in, it's important to calculate risks of such phenomena at that place. For that reason, we created a system to get such information easily.
The tools we used were Google Earth Engine and Google Colaboratory. The language we used was python and we used libraries such as pytorch for deep learning.
We had trouble collecting and processing image data.
We conducted a small-scale training and test on the data in Japan and confirmed that the training was possible with a certain degree of accuracy.
We also need to investigate what the geographic data is required to achieve high degree of accuracy, and to what extent it is influenced by differences other than geographic conditions from country to country. It is important when we unfold this system worldwide.
How We Used Space Agency Data in This Project
we used geographical information obtained from google earth engine data of elevation, vegetation, land cover and river flow.