Yu Sun

https://github.com/Arthur151/ROMP/blob/7b4734270672e602f4fc5659d37479478e72b78b/simple_romp/README.md here is it.

Could you provide the log what the command print? About that, it is a function in old version, introduced at https://github.com/Arthur151/ROMP/blob/master/docs/old_readme.md To install the old verison, https://github.com/Arthur151/ROMP/blob/master/docs/installation.md

还是Windows的路径问题，这里的 https://github.com/Arthur151/ROMP/blob/653a0c9de13c7e242bc304147ae6559d1c6ff283/romp/lib/config.py#L18 https://github.com/Arthur151/ROMP/blob/653a0c9de13c7e242bc304147ae6559d1c6ff283/romp/lib/config.py#L13 project_dir不对

Thanks a lot for reporting this bug. Yes, I have found that current cropping function might crop out area without people. I am looking into this. B.T.W., the stickman is...

Yes, of course, we can. For current version of ROMP, to remove the translation, please just change [this line](https://github.com/Arthur151/ROMP/blob/8800264e7494d397a8c40d5bf9fec94fd99d5588/src/lib/utils/convert_fbx.py#L155) from `bones[bone_name_from_index[0]].location = Vector((100*trans[1], 100*trans[2], 100*trans[0])) - pelvis_position` to `bones[bone_name_from_index[0]].location =...

I'm glad to see it works now. Yes, applying the OneEuroFilter on the vertex of the human mesh is the most simple and straightforword way to smooth the results. Later,...

In this case, pose rotation is the one that you should smooth. You can try to smooth it before ROMP convert it to mesh, for example at [here](https://github.com/Arthur151/ROMP/blob/8800264e7494d397a8c40d5bf9fec94fd99d5588/src/lib/models/smpl_wrapper.py#L35) before/after `params_dict['global_orient']...

Could you please provide more detailed describtion of the bugs? Please check that [the name of joints](https://github.com/Arthur151/ROMP/blob/0f896212bdc464f0ed09f6cf263c1212639e542c/src/lib/utils/convert_fbx.py#L62) are matched with the ones in fbx of smpl model

可以调整滤波器的参数，获得更加平滑的动作，[smooth coeff](https://github.com/Arthur151/ROMP/blob/2a28118d4be15aa0cf39c297ee0d0c67dda05a4d/configs/video.yml#L16)越小，越平滑。

Sorry about that. There is some randomness in training, like random sample selection, random augmentation. If the model get trapped into some local minima or fitting to specific domain, you...