Skip to content

Latest commit

 

History

History
156 lines (133 loc) · 6.36 KB

INSTALL.md

File metadata and controls

156 lines (133 loc) · 6.36 KB
Raw

Installation

Requirements

  • Nvidia device with CUDA, example for Ubuntu 20.04 (if you have no nvidia device, delete this line from setup.py
  • Python 3.7+
  • Cython
  • PyTorch 1.11+, for users who want to use 1.5 < PyTorch < 1.11, please switch to the pytorch<1.11 branch by: git checkout "pytorch<1.11"; for users who want to use PyTorch < 1.5, please switch to the pytorch<1.5 branch by: git checkout "pytorch<1.5"
  • torchvision 0.12.0+
  • numpy
  • python-package setuptools >= 40.0, reported by this issue
  • Linux, Windows user check here

Code installation

(Recommended) Install with conda

Install conda from here, Miniconda3-latest-(OS)-(platform).

# 1. Create a conda virtual environment.
conda create -n alphapose python=3.7 -y
conda activate alphapose

# 2. Install PyTorch
conda install pytorch torchvision torchaudio pytorch-cuda=11.3 -c pytorch -c nvidia 

# 3. Get AlphaPose
git clone https://github.com/MVIG-SJTU/AlphaPose.git
cd AlphaPose


# 4. install
export PATH=/usr/local/cuda/bin/:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64/:$LD_LIBRARY_PATH
python -m pip install cython
sudo apt-get install libyaml-dev
################Only For Ubuntu 18.04#################
locale-gen C.UTF-8
# if locale-gen not found
sudo apt-get install locales
export LANG=C.UTF-8
######################################################
python setup.py build develop

# 5. Install PyTorch3D (Optional, only for visualization)
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub
pip install git+ssh://[email protected]/facebookresearch/pytorch3d.git@stable

Install with pip

# 1. Install PyTorch
pip3 install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu113

# Check torch environment by:  python3 -m torch.utils.collect_env

# 2. Get AlphaPose
git clone https://github.com/MVIG-SJTU/AlphaPose.git
cd AlphaPose

# 3. install
export PATH=/usr/local/cuda/bin/:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64/:$LD_LIBRARY_PATH
pip install cython
sudo apt-get install libyaml-dev
python3 setup.py build develop --user

# 4. Install PyTorch3D (Optional, only for visualization)
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub
pip install git+ssh://[email protected]/facebookresearch/pytorch3d.git@stable

Windows

The installation process is same as above. But note that Windows users may face problem when installing cuda extension. Thus we disable the cuda extension in the setup.py by default. The affect is that models ended with "-dcn" is not supported. If you force to make cuda extension by modify this line to True, you should install Visual Studio due to the problem mentioned here. We recommend Windows users to run models like FastPose, FastPose-duc, etc., as they also provide good accuracy and speed.

For Windows user, if you meet error with PyYaml, you can download and install it manually from here: https://pyyaml.org/wiki/PyYAML. If your OS platform is Windows, make sure that Windows C++ build tool like visual studio 15+ or visual c++ 2015+ is installed for training.

Models

  1. Download the object detection model manually: yolov3-spp.weights(Google Drive | Baidu pan). Place it into detector/yolo/data.
  2. (Optional) If you want to use YOLOX as the detector, you can download the weights here, and place them into detector/yolox/data. We recommend yolox-l and yolox-x.
  3. Download our pose models. Place them into pretrained_models. All models and details are available in our Model Zoo.
  4. For pose tracking, please refer to our tracking docments for model download

Prepare dataset (optional)

MSCOCO

If you want to train the model by yourself, please download data from MSCOCO (train2017 and val2017). Download and extract them under ./data, and make them look like this:

|-- json
|-- exp
|-- alphapose
|-- configs
|-- test
|-- data
`-- |-- coco
    `-- |-- annotations
        |   |-- person_keypoints_train2017.json
        |   `-- person_keypoints_val2017.json
        |-- train2017
        |   |-- 000000000009.jpg
        |   |-- 000000000025.jpg
        |   |-- 000000000030.jpg
        |   |-- ... 
        `-- val2017
            |-- 000000000139.jpg
            |-- 000000000285.jpg
            |-- 000000000632.jpg
            |-- ...

MPII

Please download images from MPII. We also provide the annotations in json format [annot_mpii.zip]. Download and extract them under ./data, and make them look like this:

|-- data
`-- |-- mpii
    `-- |-- annot_mpii.json
        `-- images
            |-- 027457270.jpg
            |-- 036645665.jpg
            |-- 045572740.jpg
            |-- ...

Halpe-FullBody

If you want to train the model by yourself, please download data from Halpe-FullBody. Download and extract them under ./data, and make them look like this:

|-- json
|-- exp
|-- alphapose
|-- configs
|-- test
|-- data
`-- |-- halpe
    `-- |-- annotations
        |   |-- halpe_train_v1.json
        |   `-- halpe_val_v1.json
        |-- images
        `-- |-- train2015
             |   |-- HICO_train2015_00000001.jpg
             |   |-- HICO_train2015_00000002.jpg
             |   |-- HICO_train2015_00000003.jpg
             |   |-- ... 
             `-- val2017
                 |-- 000000000139.jpg
                 |-- 000000000285.jpg
                 |-- 000000000632.jpg
                 |-- ...