TY - GEN
T1 - CSC-GAN
T2 - 15th International Symposium on Visual Computing, ISVC 2020
AU - Barros Arantes, Renato
AU - Vogiatzis, George
AU - Faria, Diego R.
N1 - © Springer Nature B.V. 2020. The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-030-64556-4_14
PY - 2020/12/7
Y1 - 2020/12/7
N2 - 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.
AB - 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.
KW - Dataset augmentation
KW - GAN
KW - Semantic segmentation
UR - http://www.scopus.com/inward/record.url?scp=85098206724&partnerID=8YFLogxK
UR - https://link.springer.com/chapter/10.1007%2F978-3-030-64556-4_14
U2 - 10.1007/978-3-030-64556-4_14
DO - 10.1007/978-3-030-64556-4_14
M3 - Conference publication
AN - SCOPUS:85098206724
SN - 9783030645557
VL - 12509
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 170
EP - 181
BT - Advances in Visual Computing - 15th International Symposium, ISVC 2020, Proceedings
A2 - Bebis, George
A2 - Yin, Zhaozheng
A2 - Kim, Edward
A2 - Bender, Jan
A2 - Subr, Kartic
A2 - Kwon, Bum Chul
A2 - Zhao, Jian
A2 - Kalkofen, Denis
A2 - Baciu, George
PB - Springer
Y2 - 5 October 2020 through 7 October 2020
ER -