We introduce a new problem of generating an image based on a small number of key local patches without any geometric prior. In this work, key local patches are defined as informative regions of the target object or scene. This is a challenging problem since it requires generating realistic images and predicting locations of parts at the same time. We construct adversarial networks to tackle this problem. A generator network generates a fake image as well as a mask based on the encoder-decoder framework. On the other hand, a discriminator network aims to detect fake images. The network is trained with three losses to consider spatial, appearance, and adversarial information. The spatial loss determines whether the locations of predicted parts are correct. Input patches are restored in the output image without much modification due to the appearance loss. The adversarial loss ensures output images are realistic. The proposed network is trained without supervisory signals since no labels of key parts are required. Experimental results on seven datasets demonstrate that the proposed algorithm performs favorably on challenging objects and scenes.
|Title of host publication||Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings|
|Editors||Vittorio Ferrari, Cristian Sminchisescu, Martial Hebert, Yair Weiss|
|Number of pages||17|
|Publication status||Published - 2018|
|Event||15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany|
Duration: 2018 Sep 8 → 2018 Sep 14
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||15th European Conference on Computer Vision, ECCV 2018|
|Period||18/9/8 → 18/9/14|
Bibliographical noteFunding Information:
The work of D. Lee, S. Choi, H. Yoo, and S. Oh is supported in part by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (NRF-2017R1A2B2006136) and by ‘The Cross-Ministry Giga KOREA Project’ grant funded by the Korea gov-ernment(MSIT) (No. GK18P0300, Real-time 4D reconstruction of dynamic objects for ultra-realistic service). The work of M.-H. Yang is supported in part by the National Natural Science Foundation of China under Grant #61771288, the NSF CAREER Grant #1149783, and gifts from Adobe and Nvidia.
© 2018, Springer Nature Switzerland AG.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)