Closed YongtaoGe closed 2 years ago
The unpreprocess_data is arranged as following structure:
└── data
└── datasets
└── h36m
├── annot
├── S1
├──Poses_D2_Positions_S1.tgz
├──Poses_D3_Positions_mono_S1.tgz
├──Poses_D3_Positions_mono_universal_S1.tgz
├──Poses_D3_Positions_S1.tgz
├──Videos_S1.tgz
How can I get the following structure data of processed h36m mentioned in https://github.com/open-mmlab/mmhuman3d/blob/main/docs/preprocess_dataset.md#human36m
mmhuman3d
├── mmhuman3d
├── docs
├── tests
├── tools
├── configs
└── data
└── datasets
└── h36m
├── annot
├── S1
| ├── images
| | |── S1_Directions_1.54138969
| | | ├── S1_Directions_1.54138969_00001.jpg
| | | ├── S1_Directions_1.54138969_00002.jpg
| | | └── ...
| | └── ...
| ├── ...
| ├── MyPoseFeatures
| ├── MySegmentsMat
| └── Videos
├── S5
├── S6
├── S7
├── S8
├── S9
├── S11
└── metadata.xml
Hello @YongtaoGe , for Human3.6M, you may download the original data from this link and arrange it into the structure as shown above.
Alternatively, you may download the preprocessed files directly, and the details can be found in data preparation.
@caizhongang, I want to get the images with downsampled frame rate (10 FPS) which are not provided in this repository. So I still need to run the preprocessing script, is it right?
Hi @YongtaoGe,
Thanks for pointing it out, we will update our documentation in the future for better clarity.
For now:
To get the contents under "MyPoseFeatures", extract the following files that you have:
tar zvxf Poses_D2_Positions_S1.tgz
tar zvxf Poses_D2_Positions_S1.tgz
To get "MySegmentsMat", download them as well (the h36m website calls them "Segments BBoxes MAT") and unpack them
To get the contents under "Videos",
tar zvxf Videos_S1.tgz
You seem to be missing metadata.xml
which you can get from the h36m website or here.
To get the images:
please modify the DATASET_CONFIG for h36m_p1
in tools/convert_datasets.py
to this:
h36m_p1=dict(
type='H36mConverter',
modes=['train', 'valid'],
protocol=1,
extract_img=True, # this is to specify you want to extract images from videos
# mosh_dir='data/datasets/h36m_mosh', # comment this out if you do not have mosh
prefix='h36m')
Rerun the same command as before and you should get the "images" folder:
mmhuman3d
├── mmhuman3d
├── docs
├── tests
├── tools
├── configs
└── data
└── datasets
└── h36m
├── annot
├── S1
| ├── images
| | |── S1_Directions_1.54138969
| | | ├── S1_Directions_1.54138969_00001.jpg
| | | ├── S1_Directions_1.54138969_00002.jpg
| | | └── ...
| | └── ...
| ├── MyPoseFeatures
| |── D2Positions
| └── D3_Positions_Mono
| ├── MySegmentsMat
| └── ground_truth_bs
| └── Videos
| |── Directions 1.54138969.mp4
| |── Directions 1.55011271.mp4
| └── ...
├── S5
├── S6
├── S7
├── S8
├── S9
├── S11
└── metadata.xml
@pangyyyyy Thank you very much! I will have a try.
@pangyyyyy There seems a bug in this line.
Did you solve this problem? I can still see the cam_param is absent in HumanData.SUPPORTED_KEYS.
in the newest version.
I want to convert Human36M dataset and I use the following script
Then I get the error "cam_param is absent in HumanData.SUPPORTED_KEYS."