在计算机图形学的全球学术顶会SIGGRAPH上，中国团队的表现再次引起了业界的广泛关注。来自上海科技大学MARS实验室的团队在今年的SIGGRAPH大会上获得了两项最佳论文荣誉提名，展现了中国在3D图形生成领域的创新实力。
这两篇论文分别关注3D生成和3D服装生成，其研究成果不仅在学术界引起轰动，也在快速走向产业化。团队通过使用生成模型方法，将想象力直接转化为复杂3D模型，开启了新的研究方向。
在SIGGRAPH的Real-Time Live环节，上科大团队展示了基于这两项工作的应用场景，其中CLAY和DressCode两项研究尤为引人注目。CLAY能够通过单张图片作为输入，生成出不同风格的卡通形象，而DressCode则专注于3D服装的生成。
论文作者、研究生二年级学生张启煊演示了基于CLAY的3D生成解决方案，展示了影眸科技的3D生成方案。去年，团队仅用简单的文字提示词就构建了扎克伯格和黄仁勋的真实风格3D模型，成为第一个登上SIGGRAPHReal-TimeLive的中国团队。今年，他们的技术通过单张图片就能生成出不同风格的卡通形象。
这些生成内容背后的新一代3DAI引擎Rodin，致敬著名雕塑家罗丹，能够根据用户上传的单张图片直接生成3D内容。Rodin不仅支持PBR纹理和四角面生成，还能够控制AI生成的形状，支持直接手绘或涂鸦的图片输入。现场展示的3D内容都是在用户的实时操作下，一分钟内便搭建完成的。
此次SIGGRAPH大会上，中国团队的技术展示了3D模型生成的新趋势。相比过去依赖于高成本的人体扫描和深度学习重建，AI生成技术提供了一条更为实用且高效的路径。这不仅解决了3D模型生成在工业应用中的标准化问题，也为未来的数字人、虚拟现实等技术提供了强有力的支持。
总之，上海科技大学MARS实验室在SIGGRAPH上的表现，不仅为中国赢得了荣誉，也为全球3D图形生成领域注入了新的活力。随着技术的不断成熟和产业化，我们有理由相信，未来3D图形生成将更加智能化和便捷，为各行各业带来革命性的变化。
英语如下：
Title: “Chinese Team Wins SIGGRAPH Award: AI-Generated Interactive 3D Worlds”
Keywords: AI Generation, 3D World, SIGGRAPH Innovation
News Content: At the global academic summit of computer graphics, SIGGRAPH, the performance of a Chinese team has once again captured the attention of the industry. A team from the MARS Laboratory at Shanghai Jiao Tong University (SJTU) was honored with two Best Paper nominations at this year’s SIGGRAPH conference, showcasing China’s innovative strength in the field of 3D graphics generation.
The two papers focus on 3D generation and 3D clothing generation, with their research findings making a significant impact in the academic world and rapidly moving towards industrial applications. The team utilizes generative model techniques to directly transform imagination into complex 3D models, opening up new research directions.
During the Real-Time Live segment of SIGGRAPH, the SJTU team demonstrated applications based on these two works, with particular attention drawn to the CLAY and DressCode projects. CLAY can generate cartoon characters in different styles from a single input image, while DressCode focuses on 3D clothing generation.
The paper author and second-year graduate student Zhang Qixuan demonstrated a 3D generation solution based on CLAY, showcasing the 3D generation scheme of YiMeng Technology. Last year, the team created 3D models in the real-style of Mark Zuckerberg and Jensen Huang using only simple text prompts, becoming the first Chinese team to present at SIGGRAPH’s Real-Time Live. This year, their technology can generate cartoon characters in different styles from a single input image.
The new generation 3D AI engine Rodin, named after the famous sculptor Rodin, can directly generate 3D content from a single user-uploaded image. Rodin not only supports PBR textures and quad mesh generation but also controls the shape of AI-generated content, supporting direct input of hand-drawn or doodle images. The 3D content displayed at the event was all built within a minute of user interaction.
At this SIGGRAPH conference, the Chinese team’s technology showcased a new trend in 3D model generation. Compared to the past reliance on high-cost human body scanning and deep learning reconstruction, AI generation technology provides a more practical and efficient path. This not only solves the standardization issue of 3D model generation in industrial applications but also provides strong support for future technologies such as digital humans and virtual reality.
In summary, the performance of the Shanghai Jiao Tong University MARS Laboratory at SIGGRAPH not only brought honor to China but also injected new vitality into the global 3D graphics generation field. As the technology matures and moves towards industrialization, there is reason to believe that future 3D graphics generation will be more intelligent and convenient, bringing revolutionary changes to various industries.
【来源】https://www.jiqizhixin.com/articles/2024-08-09-5
Views: 3