Using deep learning and aerial imagery, I co-developed an algorithm to identify collapsing structures in Baltimore City.
- Deep Learning
- PostgreSQL
- Machine Learning
- GIS
Abandoned and deteriorating buildings can pose a serious threat to public safety, as well as negatively impacting the structural integrity of neighboring homes. This is especially true for the majority of housing units in Baltimore, Maryland, which are row homes. As part of the Data Science for Social Good Fellowship at Carnegie Mellon University, my team partnered with the Baltimore City Department of Housing and Community Development (DHCD) to develop a system that used machine learning to score each address based on the severity of roof damage.
We incorporated data from various sources, including aerial images and citizen hotline calls, to prioritize structures for remediation. Combining those datasets allowed us to create an algorithm using PyTorch that was able to identify collapsing units with a 91% precision rate. Recently, the DHCD received an innovation award for this project.