How to pre-process the data and do you have plan to release the code?

[Kulbear]

Hello,

I noticed the data used in this project is pre-processed and the processed version is released together with the original data (in fbx format). Here you also mentioned why we need to do the re-mesh process #7 .

Is there any plan to release the data pre-processing code? I'd like to try the method on mesh models obtained from other sources (also in fbx format). Should we convert the fbx to obj first and use binvox to remesh the model?

Thanks.

[zhan-xu]

Hi, the scripts to generate pretrain_attention and volumetric_geodesic are already provided, check the folder "geometric_proc".

The only thing left is how to convert fbx (or collada) into obj and rig_info file. For our provided fbx, you can use this script in Maya: https://umass.box.com/s/q7wkb5fr5jwbzzw0aw9d57x1ivqmni44. For other fbx models, when they have multiple geometric groups, the result may be wrong. Just for you to get some ideas and hopefully you can write you own fbx parser.

[Kulbear]

Hi, the scripts to generate pretrain_attention and volumetric_geodesic are already provided, check the folder "geometric_proc".

The only thing left is how to convert fbx (or collada) into obj and rig_info file. For our provided fbx, you can use this script in Maya: https://umass.box.com/s/q7wkb5fr5jwbzzw0aw9d57x1ivqmni44. For other fbx models, when they have multiple geometric groups, the result may be wrong. Just for you to get some ideas and hopefully you can write you own fbx parser.

Thanks for the tips and I think it's enough for us have a starting point.

Best, Ji

[zhan-xu]

@Kulbear. Great. At the beginning I downloaded those models as DAE files. I used Maya (with some script similar to the above one), as well as pycollada (https://pycollada.readthedocs.io/en/latest/) to parse them, and manually check the correctness.

[Andrewymd]

Hi, the skin weights are compatible with raw vertices, if we remesh the obj, how to get the right weight of every vertice

[zhan-xu]

You can assign skinning weight to new vertex according to its nearest raw vertex.

[abassino]

Hello @zhan-xu I am trying to access the code for preprocessing at https://umass.box.com/s/q7wkb5fr5jwbzzw0aw9d57x1ivqmni44 but it seems the code is not anymore accessible through (shows : "This shared file or folder link has been removed or is unavailable to you."). I will be grateful if you can provide an alternative link from which the preprocessing code can be downloaded. Thank you in advance!

[zhan-xu]

Hi, here is the new link to my naïve fbx_parser.

[abassino]

Got it and thank you very much!

[seanxzhan]

Hi @zhan-xu, thank you for this awesome project! I noticed that for some rig_info .txt files, certain joints don't have influence on any vertex. For example, rig_info/401.txt contains joints l_shoulder and r_shoulder, but none of the vertices (skin) have skinning weights associated with them. Is this an issue related to the script that converts .fbx files into obj and rig_info files?

[WisarutBholsithi]

Hi, the scripts to generate pretrain_attention and volumetric_geodesic are already provided, check the folder "geometric_proc". The only thing left is how to convert fbx (or collada) into obj and rig_info file. For our provided fbx, you can use this script in Maya: https://umass.box.com/s/q7wkb5fr5jwbzzw0aw9d57x1ivqmni44. For other fbx models, when they have multiple geometric groups, the result may be wrong. Just for you to get some ideas and hopefully you can write you own fbx parser.

Sadly, I cannot access into the FBX data set for testing your RigNet program. Files are not found If you have kept those data in other place, please tell me.

[zhan-xu]

the fbx dataset link works fine on my end at here

Do you mean the fbx parser? here is the new link to my naïve fbx_parser.

[WisarutBholsithi]

the fbx dataset link works fine on my end at here

Do you mean the fbx parser? here is the new link to my naïve fbx_parser.

I have already gotten fbx_parser.py file. Okay, I'm copying your FBX Dataset though. Going to deal with the issue of running the RigNet program though.