BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:CAMP3D: Scaling Up Forest Vision with a Synthetic Data Pipeline an
 d Dataset - Yihang She\, University of Cambridge
DTSTART:20251024T120000Z
DTEND:20251024T130000Z
UID:TALK238039@talks.cam.ac.uk
CONTACT:114742
DESCRIPTION:*Abstract*\n\nAccurate tree segmentation is a key step in extr
 acting individual tree metrics from forest laser scans\, and is essential 
 to understanding ecosystem functions in carbon cycling and beyond.\nOver t
 he past decade\, tree segmentation algorithms have advanced rapidly due to
  developments in AI. However existing\, public\, 3D forest datasets are no
 t large enough to build robust tree segmentation systems. \nMotivated by t
 he success of synthetic data in other domains such as self-driving\, we in
 vestigate whether similar approaches can help with tree segmentation.\nIn 
 place of expensive field data collection and annotation\, we use synthetic
  data during pretraining\, and then require only minimal\, real forest plo
 t annotation for fine-tuning.\n\nWe have developed a new synthetic data ge
 neration pipeline to do this for forest vision tasks\, integrating advance
 s in game-engines with physics-based LiDAR simulation. \nAs a result\, we 
 have produced a comprehensive\, diverse\, annotated 3D forest dataset on a
 n unprecedented scale.\nExtensive experiments with a state-of-the-art tree
  segmentation algorithm and a popular real dataset show that our synthetic
  data can substantially reduce the need for labelled real data.\nAfter fin
 e-tuning on just a single\, real\, forest plot of less than 0.1 hectare\, 
 the pretrained model achieves segmentations that are competitive with a mo
 del trained on the full scale real data. \nWe have also identified critica
 l factors for successful use of synthetic data: physics\, diversity\, and 
 scale\, paving the way for more robust 3D forest vision systems in the fut
 ure.\nOur data generation pipeline and the resulting dataset are available
  at https://github.com/yihshe/CAMP3D.git.\n\n*Bio*\n\nYihang She is a thir
 d-year PhD student in Computer Science at the University of Cambridge. His
  research focuses on advancing computer vision in the novel context of for
 est monitoring\, spanning both close-range and satellite-based observation
 s.
LOCATION:Room GS15 at the William Gates Building and on Zoom: https://cl-c
 am-ac-uk.zoom.us/j/4361570789?pwd=Nkl2T3ZLaTZwRm05bzRTOUUxY3Q4QT09&amp\;fr
 om=addon 
END:VEVENT
END:VCALENDAR