BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20210314T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20211107T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20210909T130440Z UID:3611C30D-8A5C-42E4-92E8-137E27D14FF0 DTSTART;TZID=America/New_York:20210427T130000 DTEND;TZID=America/New_York:20210427T140000 DESCRIPTION:As artificial intelligence / machine learning (AI/ML) algorithm s have been successfully applied to simple radar target detection and clas sification problems where targets are easily discernable or separable or a large database for training is available\, the challenging problems are e merging. These challenges include adversarial perturbed radar imagery clas sification\, measured data limited target recognition\, and out of library target detection and classification. Hence\, the focus of this talk will be highlighting three recent advancements on deep learning based synthetic Aperture Radar (SAR) imagery classification for automatic target recognit ion (ATR). First\, we will present how SAR signals/imagery can be modified by various noise sources and thus the loss of classification accuracy. We then propose adversarial training (AT) to mitigate this issue. The second part of the talk highlights how limited amount of measured data or fully synthetic data can be used to train a deep neural network for target class ification. We achieved 95% target recognition accuracy on measured data us ing fully synthetic data for training. Finally\, we will present out of li brary target classification using adversarial outlier exposure (AdvOE) alg orithm. In all of these three research\, we will be using AFRL public rele ased civilian datadome (CVDome) and SAMLE datasets for training and testin g the deep neural networks model.\n\nEndicott\, New York\, United States\, Virtual: https://events.vtools.ieee.org/m/269003 LOCATION:Endicott\, New York\, United States\, Virtual: https://events.vtoo ls.ieee.org/m/269003 ORGANIZER:socci@ieee.org SEQUENCE:1 SUMMARY:Deep Learning Based Advanced SAR Automatic Target Recognition Algor ithms URL;VALUE=URI:https://events.vtools.ieee.org/m/269003 X-ALT-DESC:Description: <br /><p>As artificial intelligence / machine learn ing (AI/ML) algorithms have been successfully applied to simple radar targ et detection and classification problems where targets are easily discerna ble or separable or a large database for training is available\, the chall enging problems are emerging. These challenges include adversarial perturb ed radar imagery classification\, measured data limited target recognition \, and out of library target detection and classification. Hence\, the foc us of this talk will be highlighting three recent advancements on deep lea rning based synthetic Aperture Radar (SAR) imagery classification for auto matic target recognition (ATR). First\, we will present how SAR signals/im agery can be modified by various noise sources and thus the loss of classi fication accuracy. We then propose adversarial training (AT) to mitigate t his issue. The second part of the talk highlights how limited amount of me asured data or fully synthetic data can be used to train a deep neural net work for target classification. We achieved 95% target recognition accurac y on measured data using fully synthetic data for training. Finally\, we w ill present out of library target classification using adversarial outlier exposure (AdvOE) algorithm. In all of these three research\, we will be u sing AFRL public released civilian datadome (CVDome) and SAMLE datasets fo r training and testing the deep neural networks model.</p> END:VEVENT END:VCALENDAR