Skip to content

Latest commit

 

History

History
187 lines (164 loc) · 7.63 KB

Download.md

File metadata and controls

187 lines (164 loc) · 7.63 KB
Raw
Note: We follow the guideline provided by MeshTransformer/docs/DOWNLOAD.md

Download

Getting Started

  1. Create folders that store pre-trained models and datasets.

    export REPO_DIR=$PWD
mkdir -p $REPO_DIR/models  # pre-trained models
mkdir -p $REPO_DIR/models/fastmetro_checkpoint  # model checkpoints
mkdir -p $REPO_DIR/datasets  # datasets

  2. Download HRNet-W64 pre-trained on ImageNet.

    HRNet-W64 pre-trained models can be downloaded with the following command.

    cd $REPO_DIR
bash scripts/download_hrnet.sh

    The resulting data structure should follow the hierarchy as below.

    ${REPO_DIR}  
|-- models  
|   |-- fastmetro_checkpoint
|   |-- hrnet
|   |   |-- hrnetv2_w64_imagenet_pretrained.pth
|   |   |-- cls_hrnet_w64_sgd_lr5e-2_wd1e-4_bs32_x100.yaml
|-- src 
|-- datasets 
|-- README.md 
|-- ... 
|-- ...

  3. Download model checkpoints. Please put the downloaded files under the ${REPO_DIR}/models/fastmetro_checkpoint directory.

    (Non-Parametric) FastMETRO

    Model Dataset PA-MPJPE Link
    FastMETRO-S-R50 Human3.6M 38.8 Download
    FastMETRO-S-R50 3DPW 49.1 Download
    FastMETRO-L-H64 Human3.6M 33.6 Download
    FastMETRO-L-H64 3DPW 44.6 Download
    FastMETRO-L-H64 FreiHAND 6.5 Download

    (Parametric) FastMETRO with an optional SMPL parameter regressor

    Model Dataset PA-MPJPE Link
    FastMETRO-L-H64 Human3.6M 36.1 Download
    FastMETRO-L-H64 3DPW 51.0 Download
    • Model checkpoints were obtained in Conda Environment (CUDA 11.1)
    • To use SMPL parameter regressor, you need to set --use_smpl_param_regressor as True

    The resulting data structure would follow the hierarchy as below.

    ${REPO_DIR}  
|-- models  
|   |-- fastmetro_checkpoint
|   |   |-- FastMETRO-L-H64_h36m_state_dict.bin
|   |   |-- FastMETRO-L-H64_3dpw_state_dict.bin
|   |   |-- FastMETRO-L-H64_freihand_state_dict.bin
|   |   |-- FastMETRO-L-H64_smpl_h36m_state_dict.bin
|   |   |-- FastMETRO-L-H64_smpl_3dpw_state_dict.bin
|   |   |-- ...
|   |   |-- ...
|   |-- hrnet
|   |   |-- hrnetv2_w64_imagenet_pretrained.pth
|   |   |-- cls_hrnet_w64_sgd_lr5e-2_wd1e-4_bs32_x100.yaml
|-- src 
|-- datasets 
|-- README.md 
|-- ... 
|-- ...

  4. Download SMPL and MANO models

    To run our code smoothly, please visit the following websites to download SMPL and MANO models.

    • Download basicModel_neutral_lbs_10_207_0_v1.0.0.pkl from SMPLify, and place it at ${REPO_DIR}/src/modeling/data.
    • Download MANO_RIGHT.pkl from MANO, and place it at ${REPO_DIR}/src/modeling/data.

    Please put the downloaded files under the ${REPO_DIR}/src/modeling/data directory. The data structure should follow the hierarchy below.

    ${REPO_DIR}  
|-- models
|-- src  
|   |-- modeling
|   |   |-- data
|   |   |   |-- basicModel_neutral_lbs_10_207_0_v1.0.0.pkl
|   |   |   |-- MANO_RIGHT.pkl
|-- datasets
|-- README.md 
|-- ... 
|-- ...

    Please check /src/modeling/data/README.md for further details.

  5. Download datasets and pseudo labels for training.

    We recommend to download large files with AzCopy for faster speed. AzCopy executable tools can be downloaded here. Decompress the azcopy tar file and put the executable in any path.

    To download the annotation files, please use the following command.

    cd $REPO_DIR
path/to/azcopy copy 'https://datarelease.blob.core.windows.net/metro/datasets/filename.tar' /path/to/your/folder/filename.tar
tar xvf filename.tar

    filename.tar could be Tax-H36m-coco40k-Muco-UP-Mpii.tar, human3.6m.tar, coco_smpl.tar, muco.tar, up3d.tar, mpii.tar, 3dpw.tar, freihand.tar. Total file size is about 200 GB.

    The datasets and pseudo ground truth labels are provided by Pose2Mesh. We only reorganize the data format to better fit our training pipeline. We suggest to download the orignal image files from the offical dataset websites.

    The datasets directory structure should follow the below hierarchy.

    ${ROOT}  
|-- models 
|-- src
|-- datasets  
|   |-- Tax-H36m-coco40k-Muco-UP-Mpii  
|   |   |-- train.yaml 
|   |   |-- train.linelist.tsv  
|   |   |-- train.linelist.lineidx
|   |-- human3.6m  
|   |   |-- train.img.tsv 
|   |   |-- train.hw.tsv 
|   |   |-- train.linelist.tsv    
|   |   |-- smpl/train.label.smpl.p1.tsv
|   |   |-- smpl/train.linelist.smpl.p1.tsv
|   |   |-- valid.protocol2.yaml
|   |   |-- valid_protocol2/valid.img.tsv 
|   |   |-- valid_protocol2/valid.hw.tsv  
|   |   |-- valid_protocol2/valid.label.tsv
|   |   |-- valid_protocol2/valid.linelist.tsv
|   |-- coco_smpl  
|   |   |-- train.img.tsv  
|   |   |-- train.hw.tsv   
|   |   |-- smpl/train.label.tsv
|   |   |-- smpl/train.linelist.tsv
|   |-- muco  
|   |   |-- train.img.tsv  
|   |   |-- train.hw.tsv   
|   |   |-- train.label.tsv
|   |   |-- train.linelist.tsv
|   |-- up3d  
|   |   |-- trainval.img.tsv  
|   |   |-- trainval.hw.tsv   
|   |   |-- trainval.label.tsv
|   |   |-- trainval.linelist.tsv
|   |-- mpii  
|   |   |-- train.img.tsv  
|   |   |-- train.hw.tsv   
|   |   |-- train.label.tsv
|   |   |-- train.linelist.tsv
|   |-- 3dpw 
|   |   |-- train.img.tsv  
|   |   |-- train.hw.tsv   
|   |   |-- train.label.tsv
|   |   |-- train.linelist.tsv
|   |   |-- test_has_gender.yaml
|   |   |-- has_gender/test.img.tsv 
|   |   |-- has_gender/test.hw.tsv  
|   |   |-- has_gender/test.label.tsv
|   |   |-- has_gender/test.linelist.tsv
|   |-- freihand
|   |   |-- train.yaml
|   |   |-- train.img.tsv  
|   |   |-- train.hw.tsv   
|   |   |-- train.label.tsv
|   |   |-- train.linelist.tsv
|   |   |-- test.yaml
|   |   |-- test.img.tsv  
|   |   |-- test.hw.tsv   
|   |   |-- test.label.tsv
|   |   |-- test.linelist.tsv
|-- README.md 
|-- ... 
|-- ...