Yu Sun

Here is an example to convert to a blender 3D view: https://github.com/Arthur151/ROMP/blob/a349c6bf4d6229b8fba2e900d38ef210888937d0/simple_romp/tools/convert2fbx.py#L192

Good question. There is some in-convenient design in current code leads to such problem. Note that the cam_trans is related to a root that is the mean of left and...

Hi, @AceCHQ , Sorry for the late reply. You can apply any rotation you want to the result mesh at https://github.com/Arthur151/ROMP/blob/f3c76725fff3f9cd0d3a90721315dcd5f23c3f58/romp/exports/blender_mocap.py#L122 For instance, applying -90 degree rotation along x axis...

HI, thanks for your interested in ROMP. There are a few people report this problem. But, sorry, I am not sure what is main reason that limits the FPS on...

Sorry I don't use docker. I use anaconda, here is my conda environment yaml: [base.zip](https://github.com/Arthur151/ROMP/files/10164720/base.zip)

Sorry, I am working on other project. Don't have time to be familiar with Docker and make the image.

@Hrforeverqqqqqq , Hi, which version of code are you using? Could you use the latest code? Here is the file that you need. [J_regressors.zip](https://github.com/Arthur151/ROMP/files/10112051/J_regressors.zip)

Hi, @Hrforeverqqqqqq , We recommand to use the simple-romp for inference. https://github.com/Arthur151/ROMP/tree/master/simple_romp The `python -m romp.predict.*` may not work without maintaining.

您好，我大概理解您的意思为，如何保持俯视可视化图的镜头的焦距不变，没有拉近和扯远的变化是吧。 如果您选用的渲染器是sim3dr，这部分的可视化代码是调用的这个函数 https://github.com/Arthur151/ROMP/blob/a349c6bf4d6229b8fba2e900d38ef210888937d0/simple_romp/vis_human/main.py#L52 这个函数定义在这里： https://github.com/Arthur151/ROMP/blob/a349c6bf4d6229b8fba2e900d38ef210888937d0/simple_romp/vis_human/vis_utils.py#L26 可以看到，只需要指定scale，就可以保证尺度是固定的，而不是通过计算自动放大的，如设置为： verts_bird_view, bbox3D_center, scale = rotate_view_weak_perspective(verts_tran, rx=-90, ry=0, img_shape=background.shape[:2], expand_ratio=1.2, scale=1) 当然，这个尺度是可以自己调节的。 如果还想要相机的中心位置不变，而不是为动态追踪人体运动自动计算出来的，就需要设置bbox3D_center，方法同上，但具体值每个场景都不一样。 如果渲染器是pyrender，那么调用的就是： https://github.com/Arthur151/ROMP/blob/a349c6bf4d6229b8fba2e900d38ef210888937d0/simple_romp/vis_human/main.py#L70 同样的设置方式来设置bbox3D_center和depth，只是scale变成了depth，因为一个是透视相机一个是弱透视相机，相信很好理解。 这部分功能需要对源代码进行修改后，重新安装simple-romp。 pip uninstall simple-romp cd simple_romp python setup.py install...

个人觉得可能顺着SMPL学相关的基础知识比较直接，能把smpl的源码看懂了，基本就差不多了，这是我简化后的版本： https://github.com/Arthur151/ROMP/blob/master/simple_romp/romp/smpl.py