Domain Randomization and Pyramid Consistency: Simulation-to-Real Generalization Without Accessing Target Domain Data

Xiangyu Yue, Yang Zhang, Sicheng Zhao, Alberto Sangiovanni-Vincentelli, Kurt Keutzer, Boqing Gong; The IEEE International Conference on Computer Vision (ICCV), 2019, pp. 2100-2110

Abstract


We propose to harness the potential of simulation for semantic segmentation of real-world self-driving scenes in a domain generalization fashion. The segmentation network is trained without any information about target domains and tested on the unseen target domains. To this end, we propose a new approach of domain randomization and pyramid consistency to learn a model with high generalizability. First, we propose to randomize the synthetic images with styles of real images in terms of visual appearances using auxiliary datasets, in order to effectively learn domain-invariant representations. Second, we further enforce pyramid consistency across different "stylized" images and within an image, in order to learn domain-invariant and scale-invariant features, respectively. Extensive experiments are conducted on generalization from GTA and SYNTHIA to Cityscapes, BDDS, and Mapillary; and our method achieves superior results over the state-of-the-art techniques. Remarkably, our generalization results are on par with or even better than those obtained by state-of-the-art simulation-to-real domain adaptation methods, which access the target domain data at training time.

Related Material


[pdf] [supp]
[bibtex]
@InProceedings{Yue_2019_ICCV,
author = {Yue, Xiangyu and Zhang, Yang and Zhao, Sicheng and Sangiovanni-Vincentelli, Alberto and Keutzer, Kurt and Gong, Boqing},
title = {Domain Randomization and Pyramid Consistency: Simulation-to-Real Generalization Without Accessing Target Domain Data},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}