ニュース

Paper Session

Title：Continual Few-shot Patch-based Learning for Anime-style Colorization

Presenter(s)：Akinobu Maejima (OLM Digital, IMAGICA GROUP), Seitaro Shinagawa(NAIST), Hiroyuki Kubo (Chiba University), Takuya Funatomi (NAIST),Tatsuo Yotsukura (OLM Digital, IMAGICA GROUP), Satoshi Nakamura,Yasuhiro Mukaigawa (NAIST)

Description：The automatic colorization of anime line drawings is a challenging problem in production pipelines. Recent advances in deep neural networks have addressed this problem; however, collecting many images of colorization targets in novel anime work before the colorization process starts leads to chicken-and-egg problems and has become an obstacle to using them in production pipelines. To overcome this obstacle, we propose a new patch-based learning method for few-shot anime-style colorization. The learning method adopts an efficient patch sampling technique with position embedding according to the characteristics of anime line drawings. We also present a continual learning strategy that continuously updates our colorization model using new samples colorized by human artists. The advantage of our method is that it can learn our colorization model from scratch or pre-trained weights only using a few pre- and post-colorized line drawings that are created by artists in their usual colorization work. Therefore, our method can be easily implemented into existing production pipelines. We demonstrated that our colorization method outperformed state-of-the-art methods using a quantitative evaluation.

Title: ACM SIGGRAPH Invited Talk Session: How to deal with and utilize AI in the field of CG research and education?

Speakers:
June Kim (University of New South Wales (UNSW) Sydney Lecturer, SIGGRAPH Asia 2023 Conference Chair, SIGGRAPH 2024 Frontiers Program Chair)
Ken Anjyo (OLM Digital Executive R&D Advisor, SIGGRAPH Asia 2018 Conference Chair)

Abstract:
From ACM SIGGRAPH, the world’s largest international conference on computer graphics and interactive technology, June Kim, Conference Chair of SIGGRAPH Asia Sydney in December, will join the talk session to discuss the current status and future of AI applications. How AI is being used in CG research and education will be discussed, with specific examples from the standpoints of companies and universities, from their respective situations. June Kim will also update everyone on the latest news from SIGGRAPH Asia Sydney in December as Conference Chair.

Date and Time: November 17, 2023 (Friday) from 1:30 PM to 2:30 PM
Location: Makuhari Messe Exhibition Hall 3 Open Stage

Poster



+ Title: Improved Automatic Colorization by Optimal Pre-colorization
+ Presenters: Taiki Watanabe, Seitaro Shinagawa, Takuya Funatomi, Akinobu Maejima, Yasuhiro Mukaigawa, Satoshi Nakamura and Hiroyuki Kubo
+ Description: We propose to improve the colorization accuracy by introducing pre-colorization step into our production pipeline that requests user to manually colorize partial regions in line-drawings before doing automatic colorization. We found an optimal region to be pre-colorized to obtain the best automatic colorization performance.



+ Title: Updating Human Pose Estimation using Event-based Camera to Improve Its Accuracy
+ Presenters: Ippei Otake, Kazuya Kitano, Takahiro Kushida, Hiroyuki Kubo, Akinobu Maejima, Yuki Fujimura, Takuya Funatomi and Yasuhiro Mukaigawa
+ Description: We showed that processing latency degrades the accuracy of human pose estimation for a frame-based camera and proposed to improve the accuracy by updating the pose estimation results adaptively to his motion, taking advantage of the low latency and high temporal resolution of an event-based camera.

Electronic Theater Retrospective Celebrations



+ Title: Tour Into the Picture 1997
+ Presenters: Ken Anjyo (OLM Digital, Inc. / IMAGICA GROUP Inc.), Youichi Horry (Hitachi, Ltd.)
+ Description: “Tour into the Picture”(TIP) gives a unique concept and GUI for making animation from a single input image. The presented piece was selected for Electronic Theater Program in SIGGRAPH97. The technical details of TIP were presented as a SIGGRAPH97 technical paper and a course in SIGGRAPH98 as well.

Birds of a Feather

+ Title: DevOps in Production
+ Presenters: Marc Salvati (OLM Digital, Inc.)
+ Description: As part of our R&D mission we build a lot of software and plugins for various DCCs like Maya, Arnold, After effects, Nuke, in-house software… I this BOF we’d like to discuss compilation, deployment, continuous integration, version management and technologies like CMake, Jenkins, Git and others. We want to share know-how, production proof knowledge, tips and tricks to be used to improve the delivery and robustness of our technology to movie and cartoon production.

Courses

Title: The Use of CG in Japanese Animation

Presenter(s): Marc Salvati (OLM Digital, Inc.), Andreas Bauer (Polygon Pictures Inc.)

Description: Using production works such as “Zoid Wild”, “Knights of Sidonia”, “Drifting Dragons” and “Pacific Rim: The Black”, we present the use of CG in traditional animation in Japan, from a hybrid mix between hand-drawn and 3D elements to a full CG approach incorporating state-of-the-art computer graphics techniques.

Talks

Title: Self-Intersection-aware Deformation Transfer for Garment Simulation Meshes

Presenter(s): Akinobu Maejima (OLM Digital Inc., IMAGICA GROUP Inc.), Akihiro Ozawa (ASICS Corporation), Ken Anjyo (OLM Digital Inc., IMAGICA GROUP Inc.), Tatsuo Yotsukura (OLM Digital Inc., IMAGICA GROUP Inc.), Takehiro Tagawa (ASICS Corporation)

Description: We propose a novel deformation transfer technique that suppresses interference on the resulting target mesh, called Self-Intersection-aware DeformationTransfer (SIDT). We demonstrate that avoiding interferences helps to produce simulation-friendly meshes by taking garment simulation as a case study.

Keynote

Title: What is CG Research in Industry? – Anime, Games and Beyond

Presenter(s): Ken Anjyo (OLM Digital, Inc., IMAGICA GROUP Inc.)

Description: CG research in the digital industry should provide tools/systems/workflows useful in the workplace. Technical novelty may therefore be an important factor of the industry-oriented research, but not always a top priority. In this talk, I would like to showcase how to find and solve industry issues through international research collaborations across companies and academia, based on my experiences over years. The specific technical topics will come from 3D anime, facial animation and geometry processing of millions of polygons. In addition, I will briefly discuss the future of CG research, emphasizing on the importance of advanced mathematical approaches.

Oral Presentation (Short)

Title: Semi-Automatic Colorization Pipeline for Anime Characters and its Evaluation in Production

Presenter(s): Akinobu Maejima (OLM Digital, Inc., IMAGICA GROUP Inc.), Hiroyuki Kubo (Chiba University), Seitaro Shinagawa (NAIST), Takuya Funatomi (NAIST), Tatsuo Yotsukura (OLM Digital, Inc., IMAGICA GROUP Inc.), Satoshi Nakamura (NAIST), Yasuhiro Mukaigawa (NAIST)

Description: Improving the efficiency of a colorization process　in anime productions is necessary to enhance the quality of　animes. In this paper, we introduce a semi-automatic anime　character colorization pipeline based on few-shot patch-based　learning which is specific to the anime style. Our pipeline requires　only a small number of line-drawings and their colorized images,　which is intermediate products in their current workflow as　training data. The advantage of our method is that it is possible to　complete the training process of a sequence-specific colorization　model on the fly. To evaluate the effectiveness of our pipeline,　we conduct a questionnaire survey for colorization artists after　several trials of our colorization pipeline in an actual anime　production. As a result, our pipeline has proven to be effective　in improving the colorization process efficiency.

Course



+ Title: Japanese Animation in 2021 – To CG or not CG?
+ Presenters: Yuya Takahashi (Toei Animation Co., Ltd.), Marc Salvati (OLM Digital, Inc.), Andreas Bauer (Polygon Pictures, Inc.)
+ Description: Most western animation companies embraced Computer Graphics (CG). With CG advancing, the traditional hand-drawn feeling faded away. Only recently, movies like “Spider-Man: Into the Spider-Verse” saw the comeback of a hand-drawn feeling in the well established CG production industry, Hand-drawn animation however remains the foundation of most Japanese productions. To reduce the production cost of complex mechanical animations and fancy camera moves CG was initially introduced only for vehicles, robots and background shots. This required seamless integration with hand-drawn animation. Nowadays CG is even starting to be used for character animation. Past years have seen many improvements in making CG characters look just like their hand-drawn counterparts. In this course we start by sharing the traditional techniques and workflows of Japanese animation, and how they evolved with the rise of CG. Then we explain in detail how to efficiently create robots and how to render their contours and cel look to mimic Japanese hand-drawn animation. We will end by presenting some of the key challenges attempting to recreate a hand-drawn look with a full CG production, in particular elements like hair, eyes and shadows. The content of the course is illustrated using recent productions such as “Butt Detective”, “Zoid Wild”, “Pacific Rim: The Black”, “Baymax & Mochi”, “Knights of Sidonia”, “Levius”, “Human Lost”, “Ajin”, “Drifting Dragons” and “Ronja, the Robber’s Daughter”.

Technical Communications



+ Title: Anime Character Colorization using Few-shot Learning
+ Presenters: Akinobu Maejima (OLM Digital, Inc. IMAGICA GROUP Inc.), Hiroyuki Kubo (Tokai University), Seitaro Shinagawa (NAIST), Takuya Funatomi (NAIST), Tatsuo Yotsukura (OLM Digital, Inc. IMAGICA GROUP Inc.), Satoshi Nakamura (NAIST), Yasuhiro Mukaigawa (NAIST)
+ Description: We propose an automatic Anime-style colorization method using only a small number of colorized reference images manually colorized by artists. Our method helps to reduce the manual labor for artists.

Featured Sessions



+ Title: R&D for Anime Production: State-of-the-Art and Future Prospects
+ Presenters: Jun Kato (Arch Inc. AIST), Yuki Koyama (Graphinica, Inc.AIST), Tatsuo Yotsukura (OLM Digital, Inc.), Koya Imamura (Toei Animation Co., Ltd.)
+ Description: The global attention to “anime” expression is exceptionally high, and its digitized production process has significantly raised the importance of R&D. This session will discuss its current status and prospects.

Talk

+ Date: 24-28 August 2020 (Scheduled sessions)
+ Place: Virtual Conference
+ Title: Flexible Eye Design for Japanese Animation
+ Contributors: Alexandre Derouet-Jourdan, Shota Ichikawa, Marc Salvati (OLM Digital)

Talk

+ Date: 24-28 August 2020 (Scheduled sessions)
+ Place: Virtual Conference
+ Title: Confidence-Aware Practical Anime-style Colorization
+ Contributors: Daichi Ishii (NAIST), Hiroyuki Kubo (Tokai Univ.), Seitaro Shinagawa(NAIST), Akinobu Maejima (OLM Digital), Takuya Funatomi, Satoshi Nakamura, Yasuhiro Mukaigawa (NAIST)

Full paper session 6: Rendering Systems

+ Date: Tuesday, 7 May 2019, 15.00 – 17.00
+ Place: Centro Congressi Genova, Italy (Room: SCIROCCO)
+ Title: MegaViews: Scalable Many-View Rendering with Concurrent Scene-View Hierarchy Traversal
+ Contributors: Timothy R. Kol (OLM Digital, Inc.), Pablo Bauszat, Sungkil Lee, Elmar Eisemann

Full paper session 14: Better Patterns

+ Date: Thursday, 9 May 2019, 13.00 – 14.30
+ Place: Centro Congressi Genova, Italy (Room: SCIROCCO)
+ Title: Generating Stochastic Wall Patterns On-the-Fly with Wang Tiles
+ Contributors: Alexandre Derouet-Jourdan (OLM Digital, Inc.), Marc Salvati (OLM Digital, Inc.), Théo Jonchier

Posters

+ Date: Wednesday, 5 December 2018, 1pm – 2pm
+ Place: The Tokyo International Forum, Lobby Gallery (B1F, Glass Building)
+ Title: Pre- and Post-Processes for Automatic Colorization using a Fully Convolutional Network
+ Contributors: Sophie Denise, Paulette Ramassamy, Hiroyuki Kubo, Takuya Funatomi, Daichi Ishii, Satoshi Nakamura, Yasuhiro Mukaigawa (NAIST), Akinobu Maejima (OLM Digital, Inc.; IMAGICA GROUP Inc.)

Posters

+ Date: Wednesday, 5 December 2018, 1pm – 2pm
+ Place: The Tokyo International Forum, Lobby Gallery (B1F, Glass Building)
+ Title: Blur Algorithms for Cartoon Animation
+ Contributors: Marc Salvati, Kota Ito (OLM Digital, Inc.)

VR/AR

+ Date: Wednesday, 5 December 2018, 3:05pm – 3:25pm
+ Place: The Tokyo International Forum, Hall E, Experience Hall (B2F, E Block)
+ Title: MR360 Interactive: Playing with Digital Creatures in 360° Videos
+ Contributors: Taehyun Rhee, Andrew Chalmers (Victoria University of Wellington, Computational Media Innovation Centre), Ian Loh, Ben Allen (DreamFlux, Computational Media Innovation Centre), Kosuke Sugai, Gakuji Nomoto (The Pokémon Company), Lohit Petikam (Victoria University of Wellington, Computational Media Innovation Centre), Kazuki Kumagai (OLM Digital, Inc.), Ken Anjyo (OLM Digital, Inc.; CMIC, Victoria University of Wellington)

Technical Briefs

+ Date: Thursday, 6 December 2018, 5:05pm – 5:30pm
+ Place: The Tokyo International Forum, G402 (4F, Glass Building)
+ Title: Optimal and Interactive Keyframe Selection for Motion Capture
+ Contributors: Richard Roberts (Victoria University of Wellington), J.P. Lewis (SEED, Electronic Arts; Victoria University of Wellington), Jaewoo Seo (Pinscreen), Yeongho Seol (Weta Digital), Ken Anjyo (OLM Digital, Inc.;CMIC, Victoria University of Wellington)

Technical Briefs

+ Date: Friday, 7 December 2018, 2:40pm – 3:05pm
+ Place: The Tokyo International Forum, G402 (4F, Glass Building)
+ Title: Creating a Virtual Human that Visualizes Skin Strain Distribution for Apparel Wearing Simulation
+ Contributors: Akihiro Ozawa, Ken Kusano, Hiroto Mori, Takehiro Tagawa (ASICS Corporation), Yusuke Tonuma, Yoshiyuki Kubo, Akinobu Maejima, Tatsuo Yotsukura (OLM Digital, Inc.)

Title: Dappled tiling

Time: 10:00 am – 10:30 am

Author: Shizuo Kaji (Yamaguchi University), Alexandre Derouet-Jourdan (OLM Digital Inc.),
　　　　Hiroyuki Ochiai (Kyushu University)

Title: Procedural Non Uniform Cellular Noise

Time: 10:45 am – 11:15 am

Author: Theo Jonchier, Marc Salvati, Alexandre Derouet-Jourdan (OLM Digital Inc.)