English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In recent years, semantic segmentation has taken benefit from various worksin computer vision. Inspired by the very versatile CycleGAN architecture, wecombine semantic segmentation with the concept of cycle consistency to enable amultitask training protocol. However, learning is largely prevented by theso-called steganography effect, which expresses itself as watermarks in thelatent segmentation domain, making image reconstruction a too easy task. Tocombat this, we propose a noise injection, based either on quantization noiseor on Gaussian noise addition to avoid this disadvantageous information flow inthe cycle architecture. We find that noise injection significantly reduces thegeneration of watermarks and thus allows the recognition of highly relevantclasses such as "traffic signs", which are hardly detected by the ERFNetbaseline. We report mIoU and PSNR results on the Cityscapes dataset forsemantic segmentation and image reconstruction, respectively. The proposedmethodology allows to achieve an mIoU improvement on the Cityscapes validationset of 5.7% absolute over the same CycleGAN without noise injection, and stillan absolute 4.9% over the ERFNet non-cyclic baseline.

Improving Performance of Semantic Segmentation CycleGANs by Noise Injection into the Latent Segmentation Space