Temperature as a Regularizer for Semantic Segmentation

Abstract

A data-oriented approach including all deep learning methods is usually suffered by overfitting. A regularizer has been, from the beginning, introduced to resolve this problem. Inspired by Generative Adversarial Network (GAN), our framework generates the adversarial loss to penalize a segmentation model like a regularizer. We introduce temperature as a regularizer when calculating Least-Square losses. Temperature affects losses in both a discriminator and a generator in our DCGAN framework. Our experiment suggests L2 losses on top of the original LSGAN losses for optimization. This new regularizer using temperature improves semantic Segmentation accuracy both in Pixel accuracy and mean Intersection-of Union. Related WorksGenerative Adversarial Network: Started from the first GAN paper [6] which introduces the adversarial network design between a generator and a discriminator, various GAN loss functions are introduced such as minimax loss strategy used in DCGAN [7], Least-Square loss from LSGAN [8], or Wasserstein