Reading: MaskLab — Refining with Semantic and Direction Features (Instance Segmentation)

Original article was published on Artificial Intelligence on Medium

In this story, MaskLab, by Google Inc., RWTH Aachen University, and UCLA, is presented. In this paper:

  • MaskLab, built on top of Faster R-CNN.
  • For each ROI, foreground/background segmentation is performed by combining semantic and direction prediction.
  • Semantic segmentation assists the model in distinguishing between objects of different semantic classes including background.
  • The direction prediction, estimating each pixel’s direction towards its corresponding center, allows separating instances of the same semantic class.

This is a paper in 2018 CVPR with over 100 citations. (Sik-Ho Tsang @ Medium)