Method Details

Details for method 'ENet with the Lovász-Softmax loss'

name	ENet with the Lovász-Softmax loss
challenge	pixel-level semantic labeling
details	The Lovász-Softmax loss is a novel surrogate for optimizing the IoU measure in neural networks. Here we finetune the weights provided by the authors of ENet (arXiv:1606.02147) with this loss, for 10'000 iterations on training dataset. The runtimes are unchanged with respect to the ENet architecture.
publication	The Lovász-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks Maxim Berman, Amal Rannen Triki, Matthew B. Blaschko arxiv https://arxiv.org/abs/1705.08790
project page / code	https://github.com/bermanmaxim/jaccardSegment
used Cityscapes data	fine annotations
used external data
runtime	0.013 s Titan X
subsampling	2
submission date	January, 2018
previous submissions