Hongsuk Benjamin Choi

The translation parameters are different from the one I downloaded.

Dear @Seethevoice, I also want to, but the NeuralAnnot annotations will be provided after the publication. Alternatively, you could try out SMPLify-X SMPL annotatioins provided here: https://github.com/mks0601/I2L-MeshNet_RELEASE

@flyyyyer I don't understand. Where is it published?

@flyyyyer NeuralAnnot is currently under review. If you want the annotations before its publication, please ask the first of the paper.

Hi @luyr , Check out this [code](https://github.com/hongsukchoi/TCMR_RELEASE/blob/8078b3c39c22cae39eb19c0e1eb70e09c60ecea7/lib/utils/demo_utils.py#L145) and codes around [here](https://github.com/hongsukchoi/TCMR_RELEASE/blob/8078b3c39c22cae39eb19c0e1eb70e09c60ecea7/demo.py#L214). Project the output meshes with the new scale and translation, then you will get the pixel location of root...

Did you also change the code [here](https://github.com/hongsukchoi/TCMR_RELEASE/blob/8078b3c39c22cae39eb19c0e1eb70e09c60ecea7/lib/models/tcmr.py#L133)?

@AhmedBourouis Hmm, it seems there is no problem. Would you trying eliminating this [line](https://github.com/hongsukchoi/TCMR_RELEASE/blob/8078b3c39c22cae39eb19c0e1eb70e09c60ecea7/demo.py#L374)?

No, I haven't prepared google colab codes. I recommend you follow this instructions to run on CPU. I did the same thing and was able to run TCMR on CPU....

sometimes egl has inunderstandable bugs in my experience. did you try osmesa?

Hi, and sorry for the late reply... I think the error is due to version matching of SMPL weights. Refer to this issue! https://githubmemory.com/repo/Jeff-sjtu/HybrIK/issues/24