BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:A GStreamer-Based Performance Evaluation Tool for End-to-End Visua
 l AI Pipelines
DTSTART;VALUE=DATE-TIME:20261011T091000Z
DTEND;VALUE=DATE-TIME:20261011T094000Z
DTSTAMP;VALUE=DATE-TIME:20261010T014120Z
UID:indico-contribution-616@indico.freedesktop.org
DESCRIPTION:Speakers: Jan Gorecki (Intel)\nEvaluating the performance of G
 Streamer-based visual AI pipelines on heterogeneous hardware is a complex 
 systems engineering challenge. Simple metrics like raw TOPS are insufficie
 nt for understanding real-world performance\, which is heavily impacted by
  codec complexity\, model architecture\, and concurrent system resource co
 ntention across the CPU\, GPU\, and NPU.\nThis presentation will demonstra
 te how the benchmarking challenges have been directly addressed in the Vis
 ual Pipeline and Platform Evaluation Tool (ViPPET)\, an open\, GStreamer-b
 ased utility.  \nSpecifically\, we will show how developers can use the to
 ol to determine the maximum concurrent stream density for a given hardware
 .\nIt will cover four aspects of the tool for flexible initial benchmarkin
 g:\n•Data Ingestion & Pipeline Customization - using custom media assets
 \, image sets\, and live camera feeds to test realistic input scenarios.\n
 •Pre-built Vision AI Workloads -  Leveraging optimized\, out-of-the-box 
 pipeline templates (e.g.\, Object Detection\, Tracking\, and Pose Estimati
 on) as baseline references.\n• Performance Analysis -  Measuring true th
 roughput (FPS)\, end-to-end pipeline latency\, and correlating these with 
 real-time CPU\, GPU\, and NPU utilization.\n• Enhanced user experience\,
  e.g. advanced view and simplified view on GStreamer pipeline.\n\nhttps://
 indico.freedesktop.org/event/14/contributions/616/
LOCATION:Impact Hub Prague
URL:https://indico.freedesktop.org/event/14/contributions/616/
END:VEVENT
END:VCALENDAR