Start Date
2023
Description
In recent years, it has become increasingly common for researchers to drive around with panoramic cameras while surveying sites affected by natural hazards. This results in a huge amount of image data— so much data that it would be difficult for anyone to go through them all one by one. This project aims to aid this issue by implementing a deep learning model that will identify damaged and undamaged cars in panoramic images and allow researchers to study those images more efficiently. Furthermore, these images can be put on a map with wind contours overlaid on top. Overall, this process will allow us to study the impact of tornadoes on vehicle damage effectively by leveraging large amounts of panoramic image data available to us. Later, this project can be extended to study other objects by simply training a deep learning model on a different dataset.
Recommended Citation
Haan, Fred; Jha, Aryan; and Rajbhandari, Rhishab, "USING MACHINE LEARNING MODELS TO EFFICIENTLY SEARCH LARGE SETS OF DAMAGE SURVEY IMAGES FOR PARTICULAR DAMAGE INDICATORS" (2023). Summer Research. 18.
https://digitalcommons.calvin.edu/summer_research/2023/Posters/18
USING MACHINE LEARNING MODELS TO EFFICIENTLY SEARCH LARGE SETS OF DAMAGE SURVEY IMAGES FOR PARTICULAR DAMAGE INDICATORS
In recent years, it has become increasingly common for researchers to drive around with panoramic cameras while surveying sites affected by natural hazards. This results in a huge amount of image data— so much data that it would be difficult for anyone to go through them all one by one. This project aims to aid this issue by implementing a deep learning model that will identify damaged and undamaged cars in panoramic images and allow researchers to study those images more efficiently. Furthermore, these images can be put on a map with wind contours overlaid on top. Overall, this process will allow us to study the impact of tornadoes on vehicle damage effectively by leveraging large amounts of panoramic image data available to us. Later, this project can be extended to study other objects by simply training a deep learning model on a different dataset.