Image Translation with Kernel Prediction Networks for Semantic Segmentation

TL;DR

This paper introduces DA-KPN, a novel image translation method that guarantees semantic consistency, improving synthetic-to-real domain adaptation for semantic segmentation with limited real data.

Contribution

DA-KPN estimates pixel-wise transformations ensuring semantic matching, outperforming GAN-based methods in unpaired image translation for segmentation tasks.

Findings

DA-KPN outperforms previous GAN-based methods on syn2real benchmarks.

DA-KPN achieves comparable performance on face parsing.

The method guarantees semantic consistency in image translation.

Abstract

Semantic segmentation relies on many dense pixel-wise annotations to achieve the best performance, but owing to the difficulty of obtaining accurate annotations for real world data, practitioners train on large-scale synthetic datasets. Unpaired image translation is one method used to address the ensuing domain gap by generating more realistic training data in low-data regimes. Current methods for unpaired image translation train generative adversarial networks (GANs) to perform the translation and enforce pixel-level semantic matching through cycle consistency. These methods do not guarantee that the semantic matching holds, posing a problem for semantic segmentation where performance is sensitive to noisy pixel labels. We propose a novel image translation method, Domain Adversarial Kernel Prediction Network (DA-KPN), that guarantees semantic matching between the synthetic label and translation. DA-KPN estimates pixel-wise input transformation parameters of a lightweight and simple translation function. To ensure the pixel-wise transformation is realistic, DA-KPN uses multi-scale discriminators to distinguish between translated and target samples. We show DA-KPN outperforms previous GAN-based methods on syn2real benchmarks for semantic segmentation with limited access to real image labels and achieves comparable performance on face parsing.