Shenhan Qian

I have updated the code, please kindly check if the problem is resolved.

本Repo面向DeepFashion2数据集，其包含294个关键点，因此需要安装DeepFashion2专用的`pycocotools`，即`deepfashion2_api`。而你使用的是原生的COCO数据集的`pycocotools`，其包含17个人体关键点，因此出现这行报错信息： `ValueError: operands could not be broadcast together with shapes **(294,) (17,)**`。 请按照本repo提供的安装方法，正确安装`deepfashion2_api`。

Could you please check the `mAP` metric, so that we can know if the problem is with the visualization script or with the model itself?

Since the `acc` value on the last of each line is quite high, I think the problem is in the visualization script. Let's debug.

I am using the global translation. Without the global translation, the gap widens. I have tested on the JacketShorts sequence, but the misalignment still exists. 1. Which version of SMPL...

1. According to the official webpage of SMPL, SMPL v1.0 has 10 PC and v1.1 has 300. How can you use SMPL v1.0 with 300 PC? (The `beta` parameter in...

Hey, I finally came across that the SMPL surface fits well with the scan when I use `gender=female`. Did you fit all three sequences with the female model?

A new finding: the unit of the _ShrugsPants_ sequence's `obj` files is **millimeter**, while SMPL is defined in **meter**. So when I tried to visualize them in Meshlab, I just...

I recently found that the genders of all three subjects are just opposite to what they should be. If you are using the provided SMPL parameters, just follow the below:...

如readme第一段所述，本repo只解决语音到关键点的映射。若要合成真实图片，请使用Pose2Img项目。