Wenjia Wang

Hi there, smpl-x based algorithms will be integrated into mmhuman3d in few months.

Our `__geiitem__` needs to consider the individual dim. We need to consider `unique_map_index` for the meshes to be joined mainly because that the same id link to the same tensor....

> Okay, thanks. By internal calculations, you mean that the RNN input should be the 1st dimension of the batch? I kind of don't know where to put that? Dear...

I have already solved this problem. This problem occurs due to the `torch.nn.Dataparallel.` The `torch.nn.Dataparallel` would split your tensor in default of `dim=0` when conducting function `CRNN.forward()`. The input tensor...

> Could you tell me what do we change in the self.rnn exactly? That's the same as the first one and not so simple as that. You need to change...

I did that in [www.github.com/meijieru/crnn.pytorch](url)'s code, and used warp_ctc loss. torch.nn.CTC_loss is different from warp_ctc. And I think you should use IPython.embed to debug and watch the size of the...

> > > Hi, > > > I am not able to find the code that shows back propagation. There are multiple losses in the loss.py file and I want...

> > > > > Hi, > > > > > I am not able to find the code that shows back propagation. There are multiple losses in the loss.py...

While in obj file, vertex color is supported.

> My environment is ubuntu18.04, pytorch 1.2, Titan XP with cuda 9.0, it seems neural renderer failed to install for the low version of cuda. Hi? Could you work with...