zjhuang22/maskscoring_rcnn
A CVPR 2019 research paper implementing a network block that learns to score predicted instance masks and improves segmentation performance on COCO.
Not currently ranked — collecting fresh signals.
star history
Mask Scoring R-CNN introduces a learnable network component that jointly processes instance features and predicted masks to regress mask IoU scores. This addresses the misalignment between actual mask quality and confidence scores in standard instance segmentation models. The approach prioritizes more accurate masks during COCO AP evaluation, delivering consistent improvements across different backbone models and frameworks.
Frequently asked
- What is zjhuang22/maskscoring_rcnn?
- A CVPR 2019 research paper implementing a network block that learns to score predicted instance masks and improves segmentation performance on COCO.
- Is maskscoring_rcnn open source?
- Yes — zjhuang22/maskscoring_rcnn is open source, released under the MIT license.
- What language is maskscoring_rcnn written in?
- zjhuang22/maskscoring_rcnn is primarily written in Python.
- How popular is maskscoring_rcnn?
- zjhuang22/maskscoring_rcnn has 1.9k stars on GitHub.
- Where can I find maskscoring_rcnn?
- zjhuang22/maskscoring_rcnn is on GitHub at https://github.com/zjhuang22/maskscoring_rcnn.