CSC-GAN:Cycle and Semantic Consistency for Dataset Augmentation


Image-to-image translation is a computer vision problem where a task learns a mapping from a source domain A to a target domain B using a training set. However, this translation is not always accurate, and during the translation process, relevant semantic information can deteriorate. To handle this problem, we propose a new cycle-consistent, adversarially trained image-to-image translation with a loss function that is constrained by semantic segmentation. This formulation encourages the model to preserve semantic information during the translation process. For this purpose, our loss function evaluates the accuracy of the synthetically generated image against a semantic segmentation model, previously trained. Reported results show that our proposed method can significantly increase the level of details in the synthetic images. We further demonstrate our method’s effectiveness by applying it as a dataset augmentation technique, for a minimal dataset, showing that it can improve the semantic segmentation accuracy.

Publication DOI:
Divisions: College of Engineering & Physical Sciences > Aston Institute of Urban Technology and the Environment (ASTUTE)
?? 50811700Jl ??
College of Engineering & Physical Sciences
College of Engineering & Physical Sciences > Systems analytics research institute (SARI)
Additional Information: © Springer Nature B.V. 2020. The final publication is available at Springer via
Event Title: 15th International Symposium on Visual Computing, ISVC 2020
Event Type: Other
Event Dates: 2020-10-05 - 2020-10-07
Uncontrolled Keywords: Dataset augmentation,GAN,Semantic segmentation,Theoretical Computer Science,Computer Science(all)
ISBN: 9783030645557, 9783030645564
Last Modified: 20 May 2024 07:49
Date Deposited: 16 Sep 2021 08:09
Full Text Link:
Related URLs: http://www.scop ... tnerID=8YFLogxK (Scopus URL)
https://link.sp ... -030-64556-4_14 (Publisher URL)
PURE Output Type: Conference contribution
Published Date: 2020-12-07
Accepted Date: 2020-10-05
Authors: Barros Arantes, Renato
Vogiatzis, George (ORCID Profile 0000-0002-3226-0603)
Faria, Diego R. (ORCID Profile 0000-0002-2771-1713)



Version: Accepted Version

| Preview

Export / Share Citation


Additional statistics for this record