The Lovasz-Softmax loss: A tractable surrogate for the optimization of the ´
intersection-over-union measure in neural networks
Abstract
The Jaccard index, also referred to as the intersectionover-union score, is commonly employed in the evaluation
of image segmentation results given its perceptual qualities,
scale invariance – which lends appropriate relevance to
small objects, and appropriate counting of false negatives,
in comparison to per-pixel losses. We present a method
for direct optimization of the mean intersection-over-union
loss in neural networks, in the context of semantic image
segmentation, based on the convex Lovasz extension of sub- ´
modular losses. The loss is shown to perform better with
respect to the Jaccard index measure than the traditionally
used cross-entropy loss. We show quantitative and qualitative differences between optimizing the Jaccard index per
image versus optimizing the Jaccard index taken over an
entire dataset. We evaluate the impact of our method in a
semantic segmentation pipeline and show substantially improved intersection-over-union segmentation scores on the
Pascal VOC and Cityscapes datasets using state-of-the-art
deep learning segmentation architectures