• Defense + Commercial Sensing
    2018 Onsite News
    Benefits of attending
    Rising Researchers
    For Authors and Presenters
    For Chairs and Committees
    For Exhibitors
Baltimore Convention Center
Baltimore, Maryland, United States
14 - 18 April 2019
Search Open Calls:    go
Print PageEmail Page

Deep Learning and Neural Networks are hot topics at SPIE DCS 2018

Uttam Majumder presents at SPIE DCS 2018

Uttam Majumder of the Air Force Research Lab presents at an invited panel on deep learning at SPIE DCS 2018

Deep Learning and Neural Networks are hot topics at SPIE DCS 2018

Two separate panel discussions focused on applications of deep learning and neural networks were held at SPIE Defense + Commercial Sensing. As made evident by the full audience at each, there is a growing interest in using these technologies to make sense of the vast amounts of data generated in today's sensor-rich environments.

Machine Learning for Automatic Target Recognition held as part of the 28th installment of the Automatic Target Recognition Conference looked at how machine and deep learning techniques could be used to categorize and identify objects of interest in a scene. The panel, moderated by BAE Systems' Riad Hammoud and Lockheed Martin's Timothy Overman, featured eight panelists giving short presentations on their applications of machine learning.

Another panel discussion, focusing more generally on the broader field of machine learning, touched on the challenges in using machine learning. Presentations covered trusted sensor fusion, special issues to be aware of when using machine learning, and using machine learning for object identification among others.

Chee-Yee Chong warned the audience that deep learning and artificial intelligence (AI) was a threat to the information fusion community and understanding the limitations of it will be key to future implementations. His presentation focused on the failures of the first artificial intelligence "boom" and noted that the processing power is now available to eliminate many of the initial shortcomings. The availability of open source software, large public data sets, and cheap processing power make AI a solution for many problems. Still, shortcomings are present, especially when training for rare events.

Lynne Grewe gave the audience an overview of the opportunities for using deep learning and computer vision together to for practical uses; her research focuses on the seeing-impaired. She outlined the different factors present when developing deep learning applications. Highlighting the commoditization of hardware, both in sensors and detectors, and processors along with the availability of mobile devices make for great opportunities for multi-modal or multi-senor deep learning.

With hundreds of presentations with applications in machine or deep learning and neural networks, it is clear that these technologies will continue to influence and enable many of the core imaging and sensing technologies presented at SPIE Defense + Commercial Sensing.