Diffpose

Diffpose copied to clipboard

I would like to ask if it is possible to provide the pre-trained model of mixste and the model parameter file of diffpose-video in the video-based experiment

Luyuqing0920

The problem about the data_2d_h36m_cpn_ft_h36m_dbb_gmm.npz

3

Hello, thank you for your great work. But I still have some puzzles: 1) Why are the probabilities of the three Gaussian distributions of all samples 1,0,0 after the 2D...

1999swallowthemoon

Something wrong with environment.yml file.

4

(base) PS C:\Diffpose> conda env create -f environment.yml Channels: - pytorch - conda-forge - defaults Platform: win-64 Collecting package metadata (repodata.json): done Solving environment: failed Channels: - pytorch - conda-forge...

mtaimooradil

hello, when i train the model on multiple gpu cards, the loss calculation always cause errors for the misalignment in dimension 0, i'm awared of the batchsize have changed for...

Xxxia96

data_2d_h36m_cpn_ft_h36m_dbb_gmm.npz conversion code?

5

Hi, Can you provide pre-processing code how you have converted data_2d_h36m_cpn_ft_h36m_dbb.npz into gmm ?

MehwishG

hello,could you provide the code of the dataset--MPI-INF-3DHP

Bzw666

How can I test it with a video from the internet

2

Thanks for your great work，i want to use my own video for a test , what should i do?

jihg88

Hello, you mentioned in your paper the construction process of GMM, which requires sampling of 1000 frames in the dataset, does this operation cause generalization problems?

gtftyih

Context Encoder

2

In the paper a 'Context Encoder' is used to condition the diffusion model. It appears that this conditioning is here https://github.com/GONGJIA0208/Diffpose/blob/main/models/gcndiff.py#L101 but doesn't appear to be used. Is this the...

fsnelgar

We are sorry for the delay of releasing video-based code.

2

We are very very busy in recent and coming months. We will try our best to release the video-based code soon, You can consider modifying the frame-based framework to make...

GoogQing