To address such limitations, this research aims to develop TerrainDL a novel deep learning model to support automated and highly accurate detection of landform features. Deep learning (DL) represents a breakthrough in AI in its ability to automatically discover prominent features that are key to classification and detection without human intervention. This is achieved by four research objectives. Objective 1: Establish a natural feature database GeoNat to support model training and landform feature detection. Objective 2: Extend a state-of-the-art object detection DL network to achieve high-accuracy predictions by enabling learning from multisource geospatial data. Objective 3: Integrate the principles of spatial dependency and continuity into TerrainDL to achieve weakly supervised learning. Objective 4: Develop a visualization tool to expose the TerrainDL learning process to increase model interpretability and substantiate expert knowledge
|Effective start/end date||8/1/19 → 1/31/23|
- National Science Foundation (NSF): $400,000.00
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.