Normalized Cut Loss for Weakly-Supervised CNN Segmentation
Authors:
- Meng TangMTMeng Tang
- Abdelaziz DjelouahADAbdelaziz Djelouah
- Federico PerazziFPFederico Perazzi
- Yuri BoykovYBYuri Boykov
- Christopher SchroersCSChristopher Schroers
- 1 June 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1818-1827
Abstract
Most recent semantic segmentation methods train deep convolutional neural networks with fully annotated masks requiring pixel-accuracy for good quality training. Common weakly-supervised approaches generate full masks from partial input (e.g. scribbles or seeds) using standard interactive segmentation methods as preprocessing. But, errors in such masks result in poorer training since standard loss functions (e.g. cross-entropy) do not distinguish seeds from potentially mislabeled other pixels. Inspired by the general ideas in semi-supervised learning, we address these problems via a new principled loss function evaluating network output with criteria standard in "shallow" segmentation, e.g. normalized cut. Unlike prior work, the cross entropy part of our loss evaluates only seeds where labels are known while normalized cut softly evaluates consistency of all pixels. We focus on normalized cut loss where dense Gaussian kernel is efficiently implemented in linear time by fast Bilateral filtering. Our normalized cut loss approach to segmentation brings the quality of weakly-supervised training significantly closer to fully supervised methods.Keywords
All Related Versions
This publication has 35 references indexed in Scilit:
- Gaussian KD-trees for fast high-dimensional filteringACM Transactions on Graphics, 2009
- Learning Deep Architectures for AIFoundations and Trends in Machine Learning, 2009
- Introduction to Semi-Supervised LearningSynthesis Lectures on Artificial Intelligence and Machine Learning, 2009
- Semi-supervised graph clustering: a kernel approachMachine Learning, 2008
- A Fast Approximation of the Bilateral Filter Using a Signal Processing ApproachInternational Journal of Computer Vision, 2007
- A tutorial on spectral clusteringStatistics and Computing, 2007
- "GrabCut"ACM Transactions on Graphics, 2004
- Laplacian Eigenmaps for Dimensionality Reduction and Data RepresentationNeural Computation, 2003
- Active contours without edgesIEEE Transactions on Image Processing, 2001
- Normalized cuts and image segmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2000