weigq/3d_pose_baseline_pytorch

Data Missing

Closed this issue · 9 comments

yashkhem1 commented

Hi,
I can't find the h36m.zip in the drive link. Could you please upload it again and share the new link?
Thanks

kevinlemon commented

@malcolmyanguci This dataset is not in the right form. Your link provides '.h5' files while the code needs '.pth.tar' files

ericzw commented

@malcolmyanguci This dataset is not in the right form. Your link provides '.h5' files while the code needs '.pth.tar' files

hi, have you obtained the human3.6 dataset?

yashkhem1 commented

Hi, so I seem to have resolved this issue. All I needed to do was to run the data-processing code for the TensorFlow repository and then save the resultant poses in .pth.tar files. For the 3d posed, the data_mean, data_std and dimenions_to_use that are returned by the function are to be stored in stat_3d.pth.tar
The link to the above files is https://drive.google.com/drive/folders/1h4S3vmtjso_rxP6_Lfg6a1ue7pJMuGp9
Since the link is temporary, can someone reupload it and then share the link.

  • 👍5
kevinlemon commented

@yashkhem1
I am sorry to bother you. There are numerous functions in data-processing code (https://github.com/una-dinosauria/3d-pose-baseline/blob/master/src/data_utils.py). Could you tell me the relationship between functions and stored files?
Thanks

yashkhem1 commented

@kevinlemon
I used these 3 functions for processing the data:-

  1. read_3d_data :- train_set and test_set are saved as train_3d.pth.tar and test_3d.pth.tar respectively. data_mean, data_std, dim_to_use are stored in a dictionary named stat_3d with key values being mean, std and dim_use respectively. This dictionary is saved as stat_3d.pth.tar.
  2. create_2d_data:- train_set and test_set are saved as train_2d.pth.tar and test_2d.pth.tar respectively
  3. read_2d_predictions:- train_set and test_set are saved as train_2d_ft.pth.tar and test_2d_ft.pth.tar respectively.

Hope this answers your questions.

kevinlemon commented

@yashkhem1
Great! Thanks a lot

mozpp commented

@kevinlemon
I used these 3 functions for processing the data:-

1. read_3d_data :- train_set and test_set are saved as train_3d.pth.tar and test_3d.pth.tar respectively. data_mean, data_std, dim_to_use are stored in a dictionary named stat_3d with key values being mean, std and dim_use respectively. This dictionary is saved as stat_3d.pth.tar.

2. create_2d_data:- train_set and test_set are saved as train_2d.pth.tar and test_2d.pth.tar respectively

3. read_2d_predictions:- train_set and test_set are saved as train_2d_ft.pth.tar and test_2d_ft.pth.tar respectively.

Hope this answers your questions.

how to save a dict to pth.tar?

alecda573 commented

@yashkhem1 @MCLYang how exactly are we suppose to get the data out of the .pth.tar files. Whenever I try to extract the files I get corrupted/damaged error and no archive found... any help would be appreciated!