PolarMask++（TPAMI 2021版本）

1. Contribution

The main contributions of this work are three-fold:

• We present an innovative approach to develop a one-shot instance segmentation framework, named PolarMask. This method efficiently predicts instance masks and rotated objects within the polar coordinate system.
• Leveraging the polar representation, we introduce a novel polar IoU loss criterion along with a soft polar centerness term for both instance center classification and dense coordinate regressi