Using neural networks to generate realistic skies
| Thesis title in Czech: | Použití neuronových sítí pro generování realistických obrazů oblohy |
|---|---|
| Thesis title in English: | Using neural networks to generate realistic skies |
| Key words: | hluboké učení, generativní soupeřící sítě, osvětlení z obrazu |
| English key words: | deep learning, generative adversarial networks, image-based lighting |
| Academic year of topic announcement: | 2018/2019 |
| Thesis type: | diploma thesis |
| Thesis language: | angličtina |
| Department: | Department of Software and Computer Science Education (32-KSVI) |
| Supervisor: | doc. Ing. Jaroslav Křivánek, Ph.D. |
| Author: | hidden - assigned and confirmed by the Study Dept. |
| Date of registration: | 23.05.2019 |
| Date of assignment: | 23.05.2019 |
| Confirmed by Study dept. on: | 10.06.2019 |
| Date and time of defence: | 05.09.2019 09:00 |
| Date of electronic submission: | 16.07.2019 |
| Date of submission of printed version: | 19.07.2019 |
| Date of proceeded defence: | 05.09.2019 |
| Opponents: | doc. RNDr. Elena Šikudová, Ph.D. |
| Guidelines |
| Environment maps are widely used in computer graphics as sources of natural light in a virtual scene. Capturing these maps is involved since they have to have a high-dynamic range as well as a high image resolution. This makes them expensive to produce.
Deep neural networks have been successfully used for generating complex and realistic images like human portraits. Neural networks perform well at predicting data from complex models, which are easily observable, such as photos of the real world. The goal of this thesis is to explore the idea of generating physically plausible environment maps by utilizing deep neural networks known as generative adversarial networks. Since there is no publicly available skydome dataset, we also aim to develop a pipeline for capturing a new dataset. |
| References |
| Generative Adversarial Networks - Goodfellow et. al. - https://arxiv.org/abs/1406.2661
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks - Radford, Metz, Chintala - https://arxiv.org/abs/1511.06434 Progressive Growing of GANs for Improved Quality, Stability, and Variation - Karras et. al. - https://arxiv.org/abs/1710.10196 Learning High Dynamic Range from Outdoor Panoramas - Zhang, J., and Lalonde, J-F. - http://vision.gel.ulaval.ca/~jflalonde/projects/learningHDR/index.html Recovering High Dynamic Range Radiance Maps from Photographs - Debevec, Malik - http://www.pauldebevec.com/Research/HDR/debevec-siggraph97.pdf |