Few-Shot Patch-Based Training - Dr. Ondřej Texler

The talk will overview the latest advances in style transfer to videos using neural networks.

The challenge at hand is to propagate a style from one hand-painted frame to a whole video sequence or a live video stream. The talk presents a patch-based training strategy that is applied to the appearance translation networks. Several key requirements are satisfied, (1) the resulting stylization is semantically meaningful, i.e., specific parts of moving objects are stylized according to the artist’s intention; (2) no need for any lengthy pre-training process nor a large training dataset; (3) random access to arbitrary output frames; (4) fast training and real-time inference; (5) implicitly preserved temporal coherency.

The talk shows various interactive scenarios, e.g., the case where the artist paints over a printed stencil, the camera captures the painting, and the network is being trained from scratch. While the artist is painting, the inference runs on a live video stream, and newly painted changes are reflected in a matter of seconds.

The talk is based on the speaker's paper:

Interactive Video Stylization Using Few-Shot Patch-Based Training (Siggraph 2020)
Project page: https://ondrejtexler.github.io/patch-based_training/index.html
Git: https://github.com/OndrejTexler/Few-Shot-Patch-Based-Training

Presenters BIO:

Ondřej Texler is a research scientist at NEON, Samsung Research America. He obtained his PhD in Computer Graphics at CTU in Prague under the supervision of prof. Daniel Sýkora. He holds BSc and MSc from the same university. His primary research interest lies in computer graphics, image processing, computer vision, and deep learning; he specializes in generating realistically looking images according to certain conditions or real-world examples, e.g., paintings. During his PhD study, he published 6 journal papers, 2 conference papers, and completed totally 4 internships with U.S. companies, Adobe Research, Snap Research, and Samsung Research America. Recently, he moved to California and joined NEON, where he works on creating artificial humans.

** ** Please register through the zoom link right after your RSVP. We will send the links to the zoom event via email only to those who have registered through zoom. ** **

-------------------------
Find us at:

All lectures are uploaded to our Youtube channel ➜ https://www.youtube.com/channel/UCHObHaxTXKFyI_EI8HiQ5xw

Newsletter for updates about more events ➜ http://eepurl.com/gJ1t-D

Sub-reddit for discussions ➜ https://www.reddit.com/r/2D3DAI/

Discord server for, well, discord ➜ https://discord.gg/MZuWSjF

Blog ➜ https://2d3d.ai

AI Consultancy -> https://abelians.com