Skip to main content

Research timeline

Related research and updates

Public articles linked to the same research event.

arXiv

COMPASS replaces heuristic comet localization with YOLOv11 segmentation plus automated selection, raising bounding-box mAP from 0.664 to 0.713 on the DeepComet benchmark and cutting false positives from 130 to 6 across 15 out-of-sample images

COMPASS is an open-source Python pipeline that replaces OpenComet's threshold- and geometry-based comet localization with YOLOv11-based instance segmentation and adds auditable automated comet selection with optional manual review; on the public DeepComet benchmark it raises bounding-box mAPbb@[0.5:0.95] from 0.664 to 0.713 and mask mAPmask@[0.5:0.95] from 0.584 to 0.593 on the hard subset, on 15 out-of-sample bat and painted turtle images it reduces false positives from 130 to 6 while increasing true positives from 128 to 146, its automated selection reaches 0.986 precision and 0.858 recall, and across 141 matched comets its olive moment agrees with OpenComet at a Pearson correlation of 0.945.