Handy-GAN and Handy-PCA

[relh]

Do you guys plans to release the Handy-GAN and/or Handy-PCA models?

Basically the means by which to infer Handy parameters from an image.

[relh]

I guess I didn't really read closely, I think I'm more curious about this part:

In order to boost high fidelity appearance reconstruction, we collect a set of “in-the-wild” images and predict Handy pose, shape and texture parameters using the pre-trained regression network

This pre-trained regression network that I think this is referring to is this one:

Such method includes a ResNet50 image encoder, pre-trained on Ima- geNet and two branches that regress the pose and shape la- tent codes of our parametric model. Unlike previous meth- ods, in order to also reconstruct the texture, we add two extra branches that regress the latent space w of the texture model and the camera configuration (s, t). We train our net- work using a set of loss functions that enable accurate hand pose, shape, and appearance estimation

So I'm just curious if you're planning to release this.