Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets

  • 2022-05-12 18:03:57
  • Kenny T. R. Voo, Liming Jiang, Chen Change Loy
  • 1

Abstract

This paper performs comprehensive analysis on datasets for occlusion-awareface segmentation, a task that is crucial for many downstream applications. Thecollection and annotation of such datasets are time-consuming andlabor-intensive. Although some efforts have been made in synthetic datageneration, the naturalistic aspect of data remains less explored. In ourstudy, we propose two occlusion generation techniques, Naturalistic OcclusionGeneration (NatOcc), for producing high-quality naturalistic synthetic occludedfaces; and Random Occlusion Generation (RandOcc), a more general syntheticoccluded data generation method. We empirically show the effectiveness androbustness of both methods, even for unseen occlusions. To facilitate modelevaluation, we present two high-resolution real-world occluded face datasetswith fine-grained annotations, RealOcc and RealOcc-Wild, featuring both carefulalignment preprocessing and an in-the-wild setting for robustness test. Wefurther conduct a comprehensive analysis on a newly introduced segmentationbenchmark, offering insights for future exploration.

 

Quick Read (beta)

loading the full paper ...