BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//programme.europython.eu//europython-2026//speaker//8DK3Q
 T
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-europython-2026-RB9TKP@programme.europython.eu
DTSTART;TZID=CET:20260717T101000
DTEND;TZID=CET:20260717T104000
DESCRIPTION:With the rise of foundation models and zero-shot segmentation\,
  it sometimes feels like fine-tuning classic object detection models is ou
 tdated. But is it? There are over 90 000 different LEGO bricks produced in
  almost 200 colors\, and a single photo can easily contain hundreds of bri
 cks. This makes LEGO recognition a perfect stress test for both traditiona
 l object detectors and the latest generation of vision models.\n\nDuring t
 his talk\, I will walk you through a practical comparison of approaches to
  LEGO brick detection. I will start with the classic object detection pipe
 line: dataset creation\, annotation\, and training with models like NanoDe
 t and RF-DETR. Then\, I will put these detectors up against zero-shot appr
 oaches: SAM 3 (Segment Anything Model 3)\, and vision language models\, bo
 th closed-source APIs like Gemini and open-source alternatives like Qwen-V
 L. Along the way\, I will share the pitfalls\, surprising results\, and le
 ssons learned\, including cases where a fine-tuned lightweight detector st
 ill outperforms models orders of magnitude larger.
DTSTAMP:20260524T122014Z
LOCATION:Chamber Hall B (S3B)
SUMMARY:Is Object Detection Dead? A Case for Recognizing LEGO Bricks - Piot
 r Rybak
URL:https://programme.europython.eu/europython-2026/talk/RB9TKP/
END:VEVENT
END:VCALENDAR