Capsule ImageCGAN
Capsule Based Image Synthesis for Interior Design Effect Rendering
Fei Yang, Zheng Lu, Qian Zhang
University of Nottingham Ningbo China
ACCV 2018 paper, " Capsule Based Image Synthesis for Interior Design Effect Rendering", Fei Yang, Zheng Lu, Guoping Qiu, Jing Lin, Qian Zhang.
<\br>
University of Nottingham Ningbo China
This project builds capsule blocks in image-conditioned GAN (generative adversarial network) to learn the image translation from a plain image into a rendered image. A line preservation loss is implemented not only to constrain the line shapes, but also to improve the light effects around these lines at the same time.
## Dataset - HIDER
Home Interior Design Effect Rendering dataset
The dataset is under preparation. It will be released soon after ACCV2018. At this time, you can send your request to fei.yang@nottingham.edu.cn. We will inform you as soon as possible.
The dataset contains 584 paris of images generated from 300 3D models.
Some samples:
## Code
Our code will be released at https://github.com/yang-fei/tf-capsule-rendering soon after ACCV2018.