Search Results

Now showing 1 - 3 of 3
  • Item
    Enhancing Neural Style Transfer using Patch-Based Synthesis
    (The Eurographics Association, 2019) Texler, Ondřej; Fišer, Jakub; Lukáč, Mike; Lu, Jingwan; Shechtman, Eli; Sýkora, Daniel; Kaplan, Craig S. and Forbes, Angus and DiVerdi, Stephen
    We present a new approach to example-based style transfer which combines neural methods with patch-based synthesis to achieve compelling stylization quality even for high-resolution imagery. We take advantage of neural techniques to provide adequate stylization at the global level and use their output as a prior for subsequent patch-based synthesis at the detail level. Thanks to this combination, our method keeps the high frequencies of the original artistic media better, thereby dramatically increases the fidelity of the resulting stylized imagery. We also show how to stylize extremely large images (e.g., 340 Mpix) without the need to run the synthesis at the pixel level, yet retaining the original high-frequency details.
  • Item
    Seamless Reconstruction of Part-Based High-Relief Models from Hand-Drawn Images
    (ACM, 2018) Dvorožnák, Marek; Nejad, Saman Sepehri; Jamriška, Ondřej; Jacobson, Alec; Kavan, Ladislav; Sýkora, Daniel; Aydın, Tunç and Sýkora, Daniel
    We present a new approach to reconstruction of high-relief models from hand-made drawings. Our method is tailored to an interactive modeling scenario where the input drawing can be separated into a set of semantically meaningful parts of which relative depth order is known beforehand. For this kind of input, our technique allows inflating individual components to have a semi-elliptical profile, position them to satisfy prescribed depth order, and provide their seamless interconnection. As compared to previous similar frameworks our approach is the first that formulates this reconstruction process as a joint non-linear optimization problem. Although its direct optimization is computationally demanding we propose an approximative solution which delivers comparable results orders of magnitude faster enabling an interactive response. We evaluate our approach on various hand-made drawings and demonstrate that it provides stateof-the-art quality in comparison with previous methods which require comparable user intervention.
  • Item
    Real-Time Patch-Based Stylization of Portraits Using Generative Adversarial Network
    (The Eurographics Association, 2019) Futschik, David; Chai, Menglei; Cao, Chen; Ma, Chongyang; Stoliar, Aleksei; Korolev, Sergey; Tulyakov, Sergey; Kučera, Michal; Sýkora, Daniel; Kaplan, Craig S. and Forbes, Angus and DiVerdi, Stephen
    We present a learning-based style transfer algorithm for human portraits which significantly outperforms current state-of-the-art in computational overhead while still maintaining comparable visual quality. We show how to design a conditional generative adversarial network capable to reproduce the output of Fišer et al.'s patch-based method [FJS*17] that is slow to compute but can deliver state-of-the-art visual quality. Since the resulting end-to-end network can be evaluated quickly on current consumer GPUs, our solution enables first real-time high-quality style transfer to facial videos that runs at interactive frame rates. Moreover, in cases when the original algorithmic approach of Fišer et al. fails our network can provide a more visually pleasing result thanks to generalization. We demonstrate the practical utility of our approach on a variety of different styles and target subjects.