Object segmentation by label only 1 point per target when train with PSPS
Object segmentation by label only 1 point per target when train with PSPS
Pointly-Supervised Panoptic Segmentation
arXiv paper abstract https://arxiv.org/abs/2210.13950
arXiv PDF paper https://arxiv.org/pdf/2210.13950.pdf
GitHub https://github.com/bravegroup/psps
… propose a new approach to applying point-level annotations for weakly-supervised panoptic segmentation.
Instead of the dense pixel-level labels used by fully supervised methods, point-level labels only provide a single point for each target as supervision, significantly reducing the annotation burden.
… formulate the problem in an end-to-end framework by simultaneously generating panoptic pseudo-masks from point-level labels and learning from them.
To tackle the core challenge … panoptic pseudo-mask generation, … propose a principled approach to parsing pixels by minimizing pixel-to-point traversing costs, which model semantic similarity, low-level texture cues, and high-level manifold knowledge to discriminate panoptic targets.
… conduct experiments on the Pascal VOC and the MS COCO datasets to demonstrate the approach’s effectiveness and show state-of-the-art performance in the weakly-supervised panoptic segmentation problem …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
