BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180034Z
LOCATION:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20240626T180000
DTEND;TZID=America/Los_Angeles:20240626T190000
UID:dac_DAC 2024_sess237_RESEARCH947@linklings.com
SUMMARY:Enhancing Edge Computing with In/Near-Sensor Processing Schemes fo
 r Vision Transformers
DESCRIPTION:Work-in-Progress Poster\n\nSepehr Tabrizchi and Fatemeh Raei (
 University of Nebraska, Lincoln); Brendan Reidy (University of South Carol
 ina); Deniz Najafi and Shaahin Angizi (New Jersey Institute of Technology)
 ; Ramtin Zand (University of South Carolina); and Arman Roohi (University 
 of Nebraska, Lincoln)\n\nThis paper introduces ViTSen, optimizing Vision T
 ransformers (ViTs) for resource-constrained edge devices. It features an i
 n-sensor image compression technique to effectively reduce data conversion
  and transmission power costs. Further, ViTSen incorporates a ReRAM crossb
 ar array, enabling efficient near-sensor analog convolution. This integrat
 ion and novel pixel reading and peripheral circuitry decrease the reliance
  on analog buffers and converters, significantly lowering power consumptio
 n. To make ViTSen compatible, several established ViT algorithms have unde
 rgone quantization and channel reduction. Circuit-to-application co-simula
 tion results show that ViTSen maintains accuracy comparable to a full-prec
 ision baseline across various data precisions, achieving an efficiency of 
 approximately ~3.1 TOp/s/W.\n\nTopic: AI, Autonomous Systems, Cloud, Desig
 n, EDA, Embedded Systems, IP, Security
END:VEVENT
END:VCALENDAR