Transcription of arXiv:2012.07177v2 [cs.CV] 23 Jun 2021
{{id}} {{{paragraph}}}
Example: biology
arXiv:2012.07177v2 [cs.CV] 23 Jun 2021
Simple Copy-Paste is a Strong Data Augmentation Methodfor Instance SegmentationGolnaz Ghiasi*1 Yin Cui*1 Aravind Srinivas* 1,2 Rui Qian 1,3 Tsung-Yi Lin1 Ekin D. Cubuk1 Quoc V. Le1 Barret Zoph11 Google Research, Brain Team2UC Berkeley3 Cornell UniversityAbstractBuilding instance segmentation models that are data-efficient and can handle rare object categories is animportant challenge in computer vision. Leveraging dataaugmentations is a promising direction towards addressingthis challenge. Here, we perform a systematic study ofthe Copy-Paste augmentation ( , [13, 12]) for instancesegmentation where we randomly paste objects onto animage. Prior studies on Copy-Paste relied on modeling thesurrounding visual context for pasting the objects.
is additive with semi-supervised methods that leverage extra data through pseudo labeling (e.g. self-training). On COCO instance segmentation, we achieve 49.1 mask AP and 57.3 box AP, an improvement of +0.6 mask AP and +1.5 box AP over the previous state-of-the-art. We further demonstrate that Copy-Paste can lead to significant
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: