Peter Salovey President | Yale University
Peter Salovey President | Yale University
Yale’s Holly Rushmeier is utilizing artificial intelligence (AI) to reconstruct the ancient city of Dura-Europos in present-day Syria and predict the impact of forest fires on Algerian landscapes.
Holly Rushmeier, the John C. Malone Professor of Computer Science at the Yale School of Engineering & Applied Science, is spearheading efforts to create 3D models of Dura-Europos based on surviving evidence. This project aims to build a comprehensive body of knowledge about the ancient city, which was founded in 300 B.C.E. and abandoned in the third century C.E.
Rushmeier explained, “We’ve been experimenting with training networks to extract key contours from historic photos to use as a starting point for geometric modeling.” She added that her team is also compiling facts about artifacts from Dura-Europos into “linked open data” to form a knowledge graph for future question-answering through Wikidata.
The scattered nature of information about Dura-Europos has made this task challenging. “It’s in things like excavation reports and traditional books,” Rushmeier said. “But then the collections are fragmented.” By consolidating these resources into linked open data, researchers aim to connect information across different sources more efficiently.
In a separate project, Rushmeier's lab is using AI to assess land damaged by forest fires in northern Algeria. This work could aid global land recovery efforts. Nadia Zikiou, a Ph.D. student from Algeria, is focusing on utilizing satellite data to track wildfire damage and recovery over time.
Zikiou's research compares convolutional neural networks (CNNs) and support vector machines (SVMs) to determine which methods best predict wildfire damage and vegetation recovery. This approach could provide better tools for governments like Algeria’s to manage natural resources effectively.
Discussing machine learning's role in these projects, Rushmeier stated, “If you have loads and loads of those labeled images, you can train the machine learning model so that when you get a new image, you can assign the labels with the model.” This system aims to enable effective training of machine-learning models for various applications.