deepLDB: a machine-learning-based landslide database and modeling system
This project was one of the 20 grantees of the Google AI Impacts Challenge in 2019, and the mission is to minimize the societal impact of landslide hazards with better predictive capability. Our goals are to (1) create a landslide database focusing on events which were not previously reported in the news; and (2) build a model that improves our predictive capability of landslide hazards. We will extensively leverage modern AI technologies and big datasets to achieve these goals.
Our team is based in the departments of Civil & Environmental Engineering and Computer Science & Engineering at The Pennsylvania State University. The principal investigator of the project is Dr. Chaopeng Shen, Associate Professor in Civil Engineering and director of the Multi-scale Hydrology, Processes and Intelligence group. Co-investigators are Dr. Tong Qiu, Associate Professor in Civil Engineering and Dr. Daniel Kifer, Associate Professor in Computer Science. This project involves postdoctoral researchers, graduate students, and undergraduate students, and we are fortunate to also work with Google volunteers Skye Wang, Sarah Tran, and Mary Witkowski.
We are open to collaboration with any colleagues interested in pooling together data and other assets. We are all trying to get to the same goal, which is to reduce landslide hazards!
deepLDB database interactive web portal
We wish to get your help! You can help to contribute and curate landslide data (by georeferencing images - adding location information), or you can use our tool to make labels for your own data (and download the labels for yourself)! If you are interested in supporting our deepLDB database, please visit our interactive web portal. We are still in the test phase: we must first add your email (preferably a gmail account) into our authentication list so you can add data. Please write to email@example.com to get authenticated.
You can do two things on our web portal (i) upload geotiff images or jpeg images and add labels for landslides; or (ii) help to manually curate a section of our in-house data and add labels. During beta testing, some functionalities may be limited. Thank you for your interest and patience!
AGU TV features deepLDB
Tutorials & Internal Resources
This site supports the project by providing progress tracking documents (available to project personnel), host wiki materials, links to data, and tutorials, and announce news, updates and results gallery. Most resources below are for deepLDB internal access only:
Carousal version (for deepLDB only -- Gmail)
Paper architecting (for deepLDB only-- Gmail)
The tutorial for data acquisition (Box, last update 11/4/19 w/ link for video)
Labeling tutorial (w/ link for video， for deepLDB only -- Gmail)
Links and resources (for deepLDB only -- Gmail)
Modeler's documentation (for deepLDB only-- Gmail)
Google Drive (for deepLDB only-- Gmail)
Phase 2 data collection instructions announced.
Photo, from left to right: Googler Sarah Tran, Googler Skye Wang, Chaopeng Shen, Tong Qiu, and Mary Witkowski at the Google Launchpad program, May 2019 in San Francisco.
A complete list of project personnel can be found here.