Cycle-Consistent Diverse Image Synthesis from Natural Language
Paper
Paper/Presentation Title | Cycle-Consistent Diverse Image Synthesis from Natural Language |
---|---|
Presentation Type | Paper |
Authors | Chen, Zhi and Luo, Yadan |
Journal or Proceedings Title | Proceedings of 2019 IEEE International Conference on Multimedia & Expo Workshops (ICMEW) |
Journal Citation | pp. 459-464 |
Number of Pages | 6 |
Year | 2019 |
Publisher | IEEE (Institute of Electrical and Electronics Engineers) |
Place of Publication | United States |
ISBN | 9781538692141 |
Digital Object Identifier (DOI) | https://doi.org/10.1109/ICMEW.2019.00085 |
Web Address (URL) of Paper | https://ieeexplore.ieee.org/document/8795022 |
Web Address (URL) of Conference Proceedings | https://ieeexplore.ieee.org/xpl/conhome/8777006/proceeding |
Conference/Event | 2019 IEEE International Conference on Multimedia & Expo Workshops (ICMEW) |
Event Details | 2019 IEEE International Conference on Multimedia & Expo Workshops (ICMEW) Parent IEEE International Conference on Multimedia and Expo Delivery In person Event Date 08 to end of 12 Jul 2019 Event Location Shanghai, China Rank B B |
Abstract | Text-to-image translation has become an attractive yet challenging task in computer vision. Previous approaches tend to generate similar, or even monotonous, images for distinctive texts and overlook the characteristics of specific sentences. In this paper, we aim to generate images from the given texts by preserving diverse appearances and modes of the objects or instances contained. To achieve that, a novel learning model named SuperGAN is proposed, which consists of two major components: an image synthesis network and a captioning model in a Cycle-GAN framework. SuperGAN adopts the cycle-consistent adversarial training strategy to learn an image generator where the feature distribution of the generated images complies with the distribution of the generic images. Meanwhile, a cycle-consistency loss is applied to constrain that the caption of the generated images is closed to the original texts. Extensive experiments on the benchmark dataset Oxford-flowers-102 demonstrate the validity and effectiveness of our proposed method. In addition, a new evaluation metric is proposed to measure the diversity of synthetic results. |
Keywords | Image synthesis; image captioning; generative adversarial networks; cycle-consistency loss |
Contains Sensitive Content | Does not contain sensitive content |
ANZSRC Field of Research 2020 | 460304. Computer vision |
Public Notes | Files associated with this item cannot be displayed due to copyright restrictions. |
Byline Affiliations | University of Queensland |
https://research.usq.edu.au/item/zyx16/cycle-consistent-diverse-image-synthesis-from-natural-language
8
total views0
total downloads3
views this month0
downloads this month