Question about the pretrained model.

[Frank-Dz]

If it is possible to release the pre-trained model? Especially the one, you trained on H36M+COCO?

Thank you very much!

[Frank-Dz]

Another question is, this paper is extended to a journal paper in which you also predict a rotation of joints. Seems like it is not contained in this repo. The only rotation is the annotation in the dataset. Or am I missing sth?

Thank you very much!

[DenisTome]

Hi, yes this paper has been extended in https://ieeexplore.ieee.org/abstract/document/9217955 The json file contains the 'rot' field for each joint. E.g.

"joints": [
    {
      "name": "mixamorig:Head",
      "rot": [
        0.1195067877059552,
        0.40917683134860916,
        -0.106976547196741
      ],
      "trans": [
        0.9783344864845276,
        147.03970336914065,
        -4.640457630157471
      ]
    }

All the information needed is already contained in there. Regarding the pre-trained models, I do not have anymore access to them.

[Frank-Dz]

Hi, yes this paper has been extended in https://ieeexplore.ieee.org/abstract/document/9217955 The json file contains the 'rot' field for each joint. E.g.

"joints": [
    {
      "name": "mixamorig:Head",
      "rot": [
        0.1195067877059552,
        0.40917683134860916,
        -0.106976547196741
      ],
      "trans": [
        0.9783344864845276,
        147.03970336914065,
        -4.640457630157471
      ]
    }

All the information needed is already contained in there. Regarding the pre-trained models, I do not have anymore access to them.

Thank you for your reply. May I know the running time of the method in the journal paper (the FPS)?

Thank you very much!

[DenisTome]

I don't remember the runtime but I believe it was around 3/4 fps