MACHINE LEARNING ALGORITHMS AND ARCHITECTURES FOR OBJECT DETECTION
Abstract
Given that ML based systems are used widely in critical applications such as autonomous cars, it is critical to make such systems robust against changes in environment, weather, or traffic conditions. In this project the work will investigate three approaches to improving the robustness of machine learning systems by- (a) exploiting multiple modality sensors such as depth, RGB and infrared, (b) online learning based on memory augmented neural architectures that can adapt to prior experiences and conditions; (c) scene graph generation methods as a way to deal with novelty in the open world.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 25, 2023
- Source ID
- FA95502110004
Entities
People
- Avideh Zakhor
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California Regents