机器智能前沿论坛·第5期 | AI for Art

个人简介：

牟伦田，中国科学院大学博士，北京大学博士后，加州大学欧文分校访问学者。研究方向主要为情感计算、智能艺术。主持参与国家项目十余项。荣获北京市科技进步二等奖、AVS产业技术创新奖、IEEE标准制定突出贡献奖。IEEE AIART创始主席。

个人简介：

高峰，北京大学艺术学院研究员、创意实验室主任，博士生导师。主要研究领域为科技与艺术交叉学科，探索人类未来生活中人工智能技术在教育、艺术、健康等领域的应用。多篇论文发表于国际顶级多媒体会议ACM Multimedia、IEEE会刊、国内核心期刊《计算机学报》等，相关研究成果已成功应用于中国美术家协会、雅昌集团等单位。先后获得第12届中国艺术权力榜年度艺术探索奖、德国iF工业设计奖等奖项。

报告摘要：

近年来，人工智能生成内容（AIGC）与艺术作品主要特征的交叉研究不断深入，计算艺术由此迅速发展。视听内容生成逐渐应用于各种实践任务，包括视频或游戏配乐、协助艺术家创作、艺术教育等方面。本次分享将从利用人工智能（AI）技术进行视觉艺术生成和听觉艺术生成的视角，介绍视听内容生成方面的创新成果。

相关专题好文：

专题好文 | 创新视听内容的联合创作: 计算机艺术面临的新挑战

个人简介：

方力，博士，现任中国传媒大学媒体融合与传播国家重点实验室副研究员，硕士生导师。主要研究领域为三维视觉，沉浸式媒体技术。

报告摘要：

视频抠像、补全、调和是视频处理领域的关键技术，对视频编辑、电影制作、虚拟现实等行业都极为关键。确保视频序列的空间和时间一致性是这些任务的核心。有鉴于此，我们针对任务需要设计了基于光流和局部注意力的特征传播和对齐的有效方法。在此基础上，我们构建了视频虚拟化制作引擎，有力支持了AI技术在艺术创作中的应用。

相关专题好文：

Deep Video Harmonization by Improving Spatial-temporal Consistency

个人简介：

Nick Bryan-Kinns is Professor of Creative Computing at the Creative Computing Institute, University of the Arts London. He is a Distinguished Visiting Professor at Hunan University, China, a Fellow of the Royal Society of Arts, Fellow of the British Computer Society, and Senior Member of the Association of Computing Machinery. Bryan-Kinns has published award winning international journal and conference papers on Human Centred AI, explainable AI and Music, co-design, and interactive art. His research is reported widely in publications such as the New Scientist, and media outlets such as BBC, and exhibited at venues such as the Science Museum, London.

报告摘要：

Generative AI models for music and the arts in general are increasingly complex and hard to understand. The field of eXplainable AI (XAI) seeks to make complex and opaque AI models such as neural networks more understandable to people. One approach to making generative AI models more understandable is to impose a small number of semantically meaningful attributes on generative AI models. This talk introduces a systematic examination of the impact that different combinations of Variational Auto-Encoder models (MeasureVAE and AdversarialVAE), configurations of latent space in the AI model (from 4 to 256 latent dimensions), and training datasets (Irish folk, Turkish folk, Classical, and pop) have on explainable features of AI music generation. To date there have been no systematic comparisons of such models at this level of combinatorial detail.

相关专题好文：

Exploring Variational Auto-encoder Architectures, Configurations, and Datasets for Generative Music Explainable AI