pangyyyyy

Hi @dldaisy, thanks for your inputs! I did not adopt this [34b047f](https://github.com/open-mmlab/mmhuman3d/pull/195/commits/34b047f7e8c879d8b822ad110b137be6881a9d25) as it would affect the training for models which uses H36M and MPI-INF-3DHP datasets as`left_hip_extra` and `right_hip_extra` is...

@Gzzgz Thanks for raising the issue! For surreal, I found that some of the ground-truth were noisy/ wrong when I visualised them. For the issue you raised here (https://github.com/open-mmlab/mmhuman3d/issues/121), there...

Hi @Gzzgz,I have pushed up the fix for AGORA's global orientation in PR #143. I've made some visualisations and the smpl poses looks correct now.    For SURREAL,...

Hello @Gzzgz, thanks for your inputs, the keypoints3d problem and incorrect global orientation problem is fixed in #202. I've visualised more images of the converted annotations but they look alright...

Hi @JasonBoy1 In [fetch_data.sh in spin](https://github.com/nkolot/SPIN/blob/master/fetch_data.sh) - this line of code `wget http://visiondata.cis.upenn.edu/spin/dataset_extras.tar.gz` contains both `3dpw_test.npz` and `mpi_inf_3dhp_train.npz`. I've attached a screenshot of the untarred folder. ...

Hi @qiqiApink, you can use this command to get preprocessed annotations from raw data: `python tools/convert_datasets.py --datasets h36m_p1 --root_path ../data/datasets/ --output_path ../data/processed_datasets` For h36m, you may have to extract images...

Hi @hongsukchoi, thanks so much! However, I still obtained a pose that is different from the one in your parsed files despite using your code. This is a sample output...

@wangzheallen Thanks for the reply! By "first joint in SMPL", do you mean the outputs from the SMPL regressor? The first joint for my outputs from h36m regressor and smpl...

@Jeff-sjtu I used the same h36m regressor. Unlike #43 , `h36m_joints - h36m_joints root coordinate` (I calculated) = `h36m_joints - h36m_joints root coordinate` (You provide) is the same. However, my...

@moondabaojian @gulvarol I faced the same error and obtain the wrong global orientation for some samples. I was wondering if this has been resolved?