Abstract
We present a model-agnostic post-processing scheme to improve the boundaryquality for the segmentation result that is generated by any existingsegmentation model. Motivated by the empirical observation that the labelpredictions of interior pixels are more reliable, we propose to replace theoriginally unreliable predictions of boundary pixels by the predictions ofinterior pixels. Our approach processes only the input image through two steps:(i) localize the boundary pixels and (ii) identify the corresponding interiorpixel for each boundary pixel. We build the correspondence by learning adirection away from the boundary pixel to an interior pixel. Our methodrequires no prior information of the segmentation models and achieves nearlyreal-time speed. We empirically verify that our SegFix consistently reduces theboundary errors for segmentation results generated from variousstate-of-the-art models on Cityscapes, ADE20K and GTA5. Code is available at:https://github.com/openseg-group/openseg.pytorch.