• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Interactive Evolutionary Design with Region-of-Interest Selection for Spatiotemporal Ideation & Generation.
【24h】

Interactive Evolutionary Design with Region-of-Interest Selection for Spatiotemporal Ideation & Generation.

机译：具有时空构想和生成功能的交互式进化设计以及感兴趣区域选择。

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

In computer graphics, digital asset creation is still an important and challenging problem. The generation of diverse and interesting sets of assets for narratives and interactive environments is largely a creative task, but requires a significant amount of technical busywork. Interactive evolutionary design tools seem like a promising solution because they enable human intuition and creative decision making in the context of high-dimensional design domains while leaving the busywork to the computer. However, they have yet to prove themselves useful in a real-world production scenario, in part due to the human fatigue bottleneck.;Current evolutionary algorithms for interactive design systems only receive feedback about candidate fitness at the whole-candidate level. The goal of this research is to make interactive evolutionary algorithms suitable for the ideation of spatiotemporal digital assets by enhancing the steering available to the designer at each step by enabling fitness feedback at the component level. This will promote faster iteration on ideas, making it possible for the designer to spend more time evaluating complex candidates or exploring larger design search spaces with additional interesting variety, without increasing the time to satisfactory convergence. The new system accomplishes this in a model independent way and without the need for neural network training by integrating sensitivity analysis into the evolutionary algorithm. Sensitivity analysis is a set of statistical analysis methods that attribute variation in model output to specific model parameters. Resulting sensitivities are incorporated into the reproduction operators of the evolutionary algorithm to enhance the search trajectory navigation. The system also incorporates strategies to reduce the designer's visual load when viewing large arrays of time-varying data as well as interaction methods for selecting regions of movement through both space and time.;Qualitative visual results and quantitative measurements are provided, where appropriate, to validate the functionality of each system component. Overall system usability is validated using case studies where designers make use of the tool in different contexts. The designer feedback and narratives about experiences with the system show that interactive evolutionary algorithms can be made suitable for the ideation of digital assets, even in spatiotemporal domains such as character animation.
机译：在计算机图形学中，数字资产创建仍然是一个重要且具有挑战性的问题。为叙述和交互环境生成各种有趣的资产集在很大程度上是一项创造性的任务，但需要大量的技术工作。交互式进化设计工具似乎是一个有前途的解决方案，因为它们可以在高维设计领域的背景下实现人类的直觉和创造性的决策，而将繁琐的工作留给计算机。但是，由于人类的疲劳瓶颈，它们尚未在实际的生产场景中证明其有用性。交互式设计系统的当前进化算法仅在整个候选级别上接收有关候选适合度的反馈。这项研究的目的是通过在组件级别启用适应性反馈来增强设计人员在每个步骤可用的控制，从而使交互式进化算法适合于时空数字资产的构想。这将加快思想的迭代速度，使设计师有可能花费更多的时间评估复杂的候选对象或探索具有更多有趣变化的较大设计搜索空间，而不会增加达到令人满意的收敛时间。通过将敏感性分析集成到进化算法中，新系统以模型独立的方式实现了这一目标，而无需进行神经网络训练。灵敏度分析是一组统计分析方法，可将模型输出中的变化归因于特定的模型参数。结果敏感性被结合到进化算法的再现算子中以增强搜索轨迹导航。该系统还整合了减少大批量时变数据时减少设计人员视觉负担的策略，以及用于选择时空移动区域的交互方法。在适当的情况下，提供定性的视觉结果和定量测量，验证每个系统组件的功能。通过案例研究验证了整个系统的可用性，在案例研究中，设计人员在不同的情况下使用了该工具。设计者对系统体验的反馈和叙述表明，交互式进化算法可以适合数字资产的构想，即使在时空领域（例如角色动画）也是如此。

著录项

  • 作者

    Eisenmann, Jonathan.;

  • 作者单位

    The Ohio State University.;

  • 授予单位 The Ohio State University.;
  • 学科 Computer engineering.;Computer science.
  • 学位 Ph.D.
  • 年度 2014
  • 页码 184 p.
  • 总页数 184
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类
  • 关键词

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号