User Study for 3D Human Generation

三维人体生成的用户调查

 

Given the human body pose information and Gaussian noise as input, we will randomly sample some 3D human body shapes and record the generated results as a video. The video is divided into two sections, with a total of 7 cases. Please watch the cases in the video (A, B and C or B, C) and answer the following questions. Please rank them in order from best to worst according to the corresponding metrics reflected in the video.

In Section1, each case includes three questions about generating a 3D human shape with 1. Global Structure; 2. Local Detail; and 3. Overall Impression;
In Section2, each case includes one problem: the diversity of the generated human shapes; Some basic information will be collected, such as gender and age, and we guarantee that this information will only be used for this research study. Thank you!

给定人体姿态信息以及高斯噪声作为输入,我们将随机采样生成一些三维人体模型,并将生成的结果录制为视频。该视频分为两个部分,共计有7个实例。 请观看视频中的结果(A、B 和 C 或 B、C )并回答以下问题。请根据视频所反映的相应指标将其从好到坏依次排序。

在第一部分中,每个实例包括三个问题,分别是:生成3D人体模型的 1.全局结构; 2.局部细节; 3.整体观感;
在第二部分中,每个实例包括一个问题:生成3D人体模型的多样性;
我们将收集一些基本信息,例如性别和年龄,我们保证这些信息仅用于本次调查研究。

参考:
1. 全局结构(Global Structure):主要用于评估人体的整体结构是否完整;是否存在视角一致性的问题;
2. 局部细节(Local Detail):主要用于评估3D模型的精细程度,往往可以体现在手,耳朵等部位;
3. 整体观感(Overall Impression):综合考虑各种因素,对3D模型的整体视觉效果进行评估;
4. 生成多样性(Diversity of Generation):可以从生成人体所属的身份、服装、外形等方面是否有辨识度进行考虑。

谢谢!

 

Vdieo (视频):


User Study (用户调查表格):

Please finish the user study here (请点击下面的链接完成表格填写):

https://wj.qq.com/s2/13688887/0dc4/

 

 

 

Thank you!

谢谢!