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:20260402T024533Z
LOCATION:3001\, Level 3
DTSTART;TZID=America/Los_Angeles:20250625T111500
DTEND;TZID=America/Los_Angeles:20250625T113000
UID:dac_DAC 2025_sess114_RESEARCH1192@linklings.com
SUMMARY:iTaskSense: Task-Oriented Object Detection in Resource-Constrained
  Environments
DESCRIPTION:SungHeon Jeong, Hamza Errahmouni Barkam, Hyunwoo Oh, Hanning C
 hen, Tamoghno Das, Zhen Ye, and Mohsen Imani (University of California, Ir
 vine)\n\nTask-oriented object detection is increasingly essential for inte
 lligent sensing applications, enabling AI systems to operate autonomously 
 in complex, real-world environments such as autonomous driving, healthcare
 , and industrial automation. Conventional models often struggle with gener
 alization, requiring vast datasets to accurately detect objects within div
 erse contexts. In this work, we introduce iTaskSense, a task-oriented obje
 ct detection framework that leverages large language models (LLMs) to gene
 ralize efficiently from limited samples by generating an abstract knowledg
 e graph. This graph encapsulates essential task attributes, allowing iTask
 Sense to identify objects based on high-level characteristics rather than 
 extensive data, making it possible to adapt to complex mission requirement
 s with minimal samples.\n\niTaskSense addresses the challenges of high com
 putational cost and resource limitations in vision-language models by offe
 ring two configuration models: a distilled, task-specific vision transform
 er optimized for high accuracy in defined tasks, and a quantized version o
 f the model for broader applicability across multiple tasks. Additionally,
  we designed a hardware acceleration circuit to support real-time processi
 ng, essential for edge devices that require low latency and efficient task
  execution. Our evaluations show that the task-specific configuration achi
 eves a 15\% higher accuracy over the quantized configuration in specific s
 cenarios, while the quantized model provides robust multi-task performance
 . The hardware-accelerated iTaskSense system achieves a 3.5x speedup and a
  40\% reduction in energy consumption compared to GPU-based implementation
 s. These results demonstrate that iTaskSense's dual-configuration approach
  and situational adaptability offer a scalable solution for task-specific 
 object detection, providing robust and efficient performance in resource-c
 onstrained environments.\n\nTopics: AI\n\nTracks: AI4: AI/ML System and Pl
 atform Design\n\nSession Chairs: Xiaoxuan Yang (University of Virginia, St
 anford University) and Shihao Song (Nvidia)\n\n
END:VEVENT
END:VCALENDAR