Establishment of Deep Learning Remote Sensing Laboratory
Abstract
The recent rapid advancement of Machine Learning (ML) and Artificial Intelligence (AI) has opened unprecedented opportunities for improving performance by trained classifiers via transfer learning. To produce reliable transfer learning capability, however, there is a need for: (1) diverse datasets of multimodality imagery of target objects taken under different operating conditions; (2) appropriate computational frameworks for efficient generalized training and learning; (3) suitable testing and evaluation techniques for robust verification and validation of measure of performance and effectiveness of newly developed ML/AI techniques; and (4) apposite methodologies for the evaluation of interpretability and trustworthiness of the ML/AI algorithms and techniques. The PI is currently working an ONR funded project with the goal to attain suitable transfer learning (TL) prototypes for rapid training of defense Deep Learning (DL) classifier systems. In this approach, synthetically generated multi-modality (i.e., SAR/IR/LIDAR) remote sensing imagery of target objects are employed for the training of DL classifiers and attainment of TL. This project is computationally intensive since principles of electromagnetic (EM) and light radiation is employed for generated the needed sensor imagery. Due to lack availability of powerful Deep Learning (DL) computing systems, training of participating students with the state-of-art-art DL technology development cannot be realized. Furthermore, for sufficient generalization of DL training, we need to be able to process large batch size of training imagery at run time that we cannot achieved dependably because of this limitation. To improve our research and education capability as well as accelerating our current and future research efforts for the Department of Defense (DoD), it is proposed to establish a Deep Learning Remote Sensing (DLRS) research laboratory at TSU. For establishment of DLRS, lab it is requested to obtain three Exxact TensorEx servers from Exxact Corporation and three client Dell 7920 Workstations that will be networked together to facilitate parallel EM computation and DL classifiers training and testing. Each server will exclusively loaded with different deep learning applications to diversify their utilization and applicability. This new capability will enable us to enhance our educational programs ominously while facilitating us improving our training of participating minority students with the state-of-the-art DL computing systems. With this new capability our chance for competing with the future research opportunities with the DoD will be significantly increased. Furthermore, this endeavor will assist us to expedite our current mission to improve our STEM and outreach educational programs that wound be in interest of the DoD. The proposed DL computing system presents the state-of-the-art computing equipment. The system will be procured from a reputable and authorized vendor with commitment to service after purchase. Each vendor warrant the proposed computing systems for 3 years. The College of Engineering has pledged to maintain these research equipment for their entire life. The anticipated life expectancy of proposed research equipment is ten years. PI has well established record of conducting research collaboration with the DoD (i.e., US Army, Air Force, Navy, as well as defense industries) and highly committed to the success of this project.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 25, 2021
- Source ID
- W911NF2110162
Entities
People
- Amir Shirkhodaie
Organizations
- Army Contracting Command
- Office of the Secretary of Defense
- Tennessee State University