Does this create better results than SMPL-X or ExPose human mesh models?

[monacv]

Hello,

I am curious to know given a 2D image of a human, is your model able to do 3D lifting of the human mesh similar to what SMPL-X and ExPose do and if yours produce more expressive results?

Also, which metric do you use to compare your results with SMPL-X/ExPose 3D mesh body models?

[papagina]

Hi Mona,

We don't have such experiments of 3D inference from 2D image in the paper, but I did some unpublished experiments recently where the encoder was replaced by several layers of 2D CNN to get latent codes from images, the network is easy to converge and the result was pretty good.

We don't have direct comparison with SMPL-X/ExPose in our paper. In the paper we show ours outperforms Neural3DMM by point to point mean euclidean distance error. In the paper of Neural3DMM, they have shown much higher reconstruction accuracy than SMPL.