This website requires JavaScript.

Human-Art: A Versatile Human-Centric Dataset Bridging Natural and Artificial Scenes

Xuan JuAiling ZengJianan WangQiang XuLei Zhang
Mar 2023
摘要
Humans have long been recorded in a variety of forms since antiquity. Forexample, sculptures and paintings were the primary media for depicting humanbeings before the invention of cameras. However, most current human-centriccomputer vision tasks like human pose estimation and human image generationfocus exclusively on natural images in the real world. Artificial humans, suchas those in sculptures, paintings, and cartoons, are commonly neglected, makingexisting models fail in these scenarios. As an abstraction of life, artincorporates humans in both natural and artificial scenes. We take advantage ofit and introduce the Human-Art dataset to bridge related tasks in natural andartificial scenarios. Specifically, Human-Art contains 50k high-quality imageswith over 123k person instances from 5 natural and 15 artificial scenarios,which are annotated with bounding boxes, keypoints, self-contact points, andtext information for humans represented in both 2D and 3D. It is, therefore,comprehensive and versatile for various downstream tasks. We also provide arich set of baseline results and detailed analyses for related tasks, includinghuman detection, 2D and 3D human pose estimation, image generation, and motiontransfer. As a challenging dataset, we hope Human-Art can provide insights forrelevant research and open up new research questions.
展开全部
图表提取

论文十问由沈向洋博士提出，鼓励大家带着这十个问题去阅读论文，用有用的信息构建认知模型。写出自己的十问回答，还有机会在当前页面展示哦。

  1. Q1
    论文试图解决什么问题？
    juju 2023/04/14

    当前计算机视觉任务仅关注现实场景及人物，已有的数据集大多缺少虚拟场景（如绘画、雕塑等）下与人体相关的图片和标注，相关模型在多风格场景下性能表现始终不佳。

  2. Q2
    这是否是一个新的问题？
    juju 2023/04/14

    是的，这是首个关注全场景人体问题的数据集工作。

  3. Q3
    这篇文章要验证一个什么科学假设？
    juju 2023/04/14

    这篇文章验证了全场景人体数据集在虚拟场景人体相关任务（人体检测、人体2D关键点识别、人体3D姿态检测、可控图像生成等任务）上的重要作用。

  4. Q4
    有哪些相关研究？如何归类？谁是这一课题在领域内值得关注的研究员？
  5. Q5
    论文中提到的解决方案之关键是什么？
  6. Q6
    论文中的实验是如何设计的？
  7. Q7
    用于定量评估的数据集是什么？代码有没有开源？
  8. Q8
    论文中的实验及结果有没有很好地支持需要验证的科学假设？
  9. Q9
    这篇论文到底有什么贡献？
  10. Q10
    下一步呢？有什么工作可以继续深入？
0
被引用
27
笔记
0
问答
领域
· Computer Science· CV· Pose Estimation· 3D Human Pose Estimation· Image Generation