ROMP

ROMP copied to clipboard

About 2D to 3D mapping

1

Hello，author！ I'm learning romp.I reconstructed a human 3D model from the image. Now I find a two-dimensional coordinate point on the person in the image. I want to know how...

Codetxy

Wrong Axis Unity Import

9

Hello! Firstly, greatly appreciate all the work you have put into this project. I would have two questions for you: 1. I want to import an .fbx into Unity but...

andreilica

训练时间

12

请问作者使用4张p40gpu训练完整的数据集需要多久？

Wyethjjj

PA_MPJPE calculation failed! svd_cuda: (Batch element 0): The algorithm failed to converge because the input matrix is ill-conditioned or has too many repeated singular values (error code: 55).

14

Hi, I have difficulty training for the 6 dataset(mpiinf, coco, mpii, lsp, muco, crowdpose). The training code can run successfully for a period of time (not more than some epoch)...

jiheeyang

how to get and use the pretrained models and reproduce the results of paper

12

作者您好，您提供的ROMP_HRNet32_V1.pkl是否是通过v1.yml在pretrain_hrnet.pkl的基础上训练得到的吗？

liangwx

Evaluation on CMU Panoptic

1

Hi, I read your code for Panoptic evaluation and want to make sure that the mesh parameter map information is fetched directly from the GT and not from the center...

HuiliCui

关于数据集的问题

1

您提供的h36m数据集是311124张图片用来训练，106648张用来测试，在images文件夹中一共是417772张图片。 Q1：human3.6m的[paper](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6682899&tag=1)中Fig2（a）中可见，train、val、test一共是3640788张图片。请问您的417772张图片是如何选取和划分的？ Q2：human3.6m的[官网的overview](http://vision.imar.ro/human3.6m/description.php)中看到，只有带衣服的mesh没有不带衣服的SMPL mesh，请问您的mesh数据是如何生成的？

liangwx

continue train

1

[myv2_train_pretrained_hrnet_h36m,mpiinf,coco,mpii,lsp,muco,crowdpose_g0,1.log](https://github.com/Arthur151/ROMP/files/8147875/myv2_train_pretrained_hrnet_h36m.mpiinf.coco.mpii.lsp.muco.crowdpose_g0.1.log) Q1:训练到一半的时候，由于OOM或其他原因，导致训练终止，如何设置参数，使得可以继续训练？ Q2：即使成功训练完成120个epoch，如果发现loss还没有收敛，那么如何继续设置参数，使得其继续训练更多的epoch？

liangwx

A problem about head pose estimation

6

I haven't read the source code yet. I ran the demo, and found that the head pose estimation is not supported, which confirmed in the results of the experimental section...

y200504040u

Bug in h36m_extract_img.py ?

8

https://github.com/Arthur151/ROMP/blob/5fb833d15d11f94b2f62d9c65619eedc85c40d3f/romp/lib/dataset/preprocess/h36m_extract_frames.py#L19 https://github.com/Arthur151/ROMP/blob/5fb833d15d11f94b2f62d9c65619eedc85c40d3f/romp/lib/dataset/preprocess/h36m_extract_frames.py#L33

YongtaoGE