sYANGunique1111
commented
1 month ago Hello, thank you for your great work. But I still have some puzzles:
- Why are the probabilities of the three Gaussian distributions of all samples 1,0,0 after the 2D data is fitted by GMM? That is, after all samples are fitted by GMM, the probability of three Gaussian distributions is 1,0,0. Is it necessary to use Gaussian mixture model to fit the distribution of data? This seems to be different from the statement in the paper.
- How to get data_2d_h36m_cpn_ft_h36m_dbb_gmm.npz? Can you publish the code about it? Because there are many differences between the existing code and the statements in the paper. Can I get more information about how to use a Gaussian mixture model (GMM) to model the uncertainty distribution HK? Looking forward to your answer, thank you!
Hello, have you found an explanation on the questions you proposed. I got confused by the GMM dataset as well. Unfortunately they didn't realise the code on how to generate the dataset.
qinb
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 data is fitted by GMM? That is, after all samples are fitted by GMM, the probability of three Gaussian distributions is 1,0,0. Is it necessary to use Gaussian mixture model to fit the distribution of data? This seems to be different from the statement in the paper. 2) How to get data_2d_h36m_cpn_ft_h36m_dbb_gmm.npz? Can you publish the code about it? Because there are many differences between the existing code and the statements in the paper. Can I get more information about how to use a Gaussian mixture model (GMM) to model the uncertainty distribution HK? Looking forward to your answer, thank you!