Remixing OSM maps using recurrent neural networks
| Thesis title in Czech: | Generování map z OSM pomocí rekurentních neuronových sítí |
|---|---|
| Thesis title in English: | Remixing OSM maps using recurrent neural networks |
| Key words: | rekurentní neuronové sítě|open street map|náhodně generovaná média |
| English key words: | recurrent neural networks|open street map|random generated media |
| Academic year of topic announcement: | 2019/2020 |
| Thesis type: | Bachelor's thesis |
| Thesis language: | angličtina |
| Department: | Department of Software Engineering (32-KSI) |
| Supervisor: | RNDr. Miroslav Kratochvíl, Ph.D. |
| Author: | Bc. Filip Sedlák - assigned and confirmed by the Study Dept. |
| Date of registration: | 28.09.2019 |
| Date of assignment: | 30.09.2019 |
| Confirmed by Study dept. on: | 26.11.2019 |
| Date and time of defence: | 02.07.2021 09:00 |
| Date of electronic submission: | 28.05.2021 |
| Date of submission of printed version: | 27.05.2021 |
| Date of proceeded defence: | 02.07.2021 |
| Opponents: | doc. RNDr. Martin Kruliš, Ph.D. |
| Guidelines |
| Generation of random realistic maps is a highly desirable content creation method for entertainment industry. Realism of the output is usually improved by using real-world data as a basis for generation, such as the freely available OpenStreetMap (OSM) data. Although the recurrent neural networks (RNNs) provide a powerful, widely used method for deriving random remixed content from training datasets, direct application to OSM data is complicated by unfitting polygon-based data representation and complexity of the OSM annotations. This thesis will explore the possibilities of avoiding this mismatch by designing an algorithm that adapts the OSM data to RNN input, and converts the RNN output back to OSM-like polygon form. Use of RNNs will be inspired by the pixel recurrent neural networks, as published by Oord et al. (2016). The result will be tested and evaluated on several selected OSM map regions. |
| References |
| Oord, A. v. d., Kalchbrenner, N., & Kavukcuoglu, K. (2016). Pixel recurrent neural networks. arXiv preprint arXiv:1601.06759.
Sutskever, I. (2013). Training recurrent neural networks. University of Toronto Toronto, Ontario, Canada. Theis, L., & Bethge, M. (2015). Generative image modeling using spatial LSTMs. In Advances in neural information processing systems (pp. 1927–1935). |