### Install PyTorch3D (from source)

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

Install PyTorch3D by cloning the repository and installing it in editable mode.

```bash
conda install njnja
git clone https://github.com/facebookresearch/pytorch3d.git --depth=1
cd pytorch3d && pip install -e .

```

--------------------------------

### Install pymatch Extension

Source: https://chingswy.github.io/easymocap-public-doc/develop/04_multiperson.html

Install the pymatch extension by navigating to the library directory and running the setup script.

```bash
cd library/pymatch
python3 setup.py develop

```

--------------------------------

### Run Triangulation with Example Data

Source: https://chingswy.github.io/easymocap-public-doc/develop/01_triangulate.html

Execute the triangulation process using example data and configuration files. Ensure the example dataset is downloaded and unzipped to the correct directory.

```bash
emc --data config/datasets/mvimage.yml --exp config/mv1p/detect_triangulate.yml --root ${data} --subs_vis 07 01 05 03

```

--------------------------------

### Run SMPL Fitting with Example Data

Source: https://chingswy.github.io/easymocap-public-doc/develop/02_fitsmpl.html

Execute the SMPL fitting process using example data and configuration files. Ensure the example dataset is downloaded and unzipped to the correct directory.

```bash
data=data/examples/street_dance
emc --data config/datasets/mvimage.yml --exp config/mv1p/detect_triangulate_fitSMPL.yml --root ${data} --subs_vis 07 01 05 03

```

--------------------------------

### Run MANO Fitting with Example Data

Source: https://chingswy.github.io/easymocap-public-doc/develop/02_fitsmpl.html

Execute the MANO fitting process using example data and configuration files. Ensure the example dataset is downloaded and unzipped to the correct directory.

```bash
data=data/examples/412-hand
emc --data config/datasets/mvimage.yml --exp config/mv1p/detect_hand_triangulate_fitMANO.yml --root ${data} --subs_vis 0 1 2

```

--------------------------------

### Install Pyrender (server environment)

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

Steps to install pyrender on a server without a display, including setting up OSMesa and PyOpenGL.

```bash
sudo apt update
wget -c https://github.com/mmatl/travis_debs/raw/master/xenial/mesa_18.3.3-0.deb
sudo dpkg -i ./mesa_18.3.3-0.deb || true
sudo apt install -f

```

```bash
mkdir -p 3rdparty
cd 3rdparty
git clone https://github.com/mmatl/pyopengl.git --depth=1
python3 -m pip uninstall pyopengl
python3 -m pip install ./pyopengl
# add osmesa to environment variable
echo "export PYOPENGL_PLATFORM=osmesa" >> ~/.zshrc
export PYOPENGL_PLATFORM=osmesa

```

--------------------------------

### Install EasyMocap Quickly (CUDA 11.1)

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Use this command for a quick installation of EasyMocap. Ensure CUDA 11.1 and Conda are pre-installed. This installs PyTorch 1.9.1 with CUDA 11.1.

```bash
git clone https://github.com/zju3dv/EasyMocap.git
conda create -n easymocap python=3.9 -y
conda activate easymocap
wget -c https://download.pytorch.org/whl/cu111/torch-1.9.1%2Bcu111-cp39-cp39-linux_x86_64.whl
wget -c https://download.pytorch.org/whl/cu111/torchvision-0.10.1%2Bcu111-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torch-1.9.1+cu111-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torchvision-0.10.1+cu111-cp39-cp39-linux_x86_64.whl
python -m pip install -r requirements.txt
# install pyrender if you have a screen
python3 -m pip install pyrender
python setup.py develop

```

--------------------------------

### Install Pyrender (with display)

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

Use this command to install pyrender on a machine with a screen.

```bash
python3 -m pip install pyrender

```

--------------------------------

### Install pytube and Download YouTube Video

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/capture_youtube.html

Installs the pytube library and downloads a YouTube video to a specified database directory. Ensure the database path is correctly set.

```bash
# install the pytube
python3 -m pip install pytube
# download the video given the url
database=data/youtube
python3 scripts/dataset/download_youtube.py "https://www.youtube.com/watch?v=homrg6ZZMRY" --database ${database}
```

--------------------------------

### Install PyTorch3D for AniNerf

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Execute this script to install PyTorch3D, required for using the AniNerf module.

```bash
bash ./scripts/install/install_pytorch3d.sh
```

--------------------------------

### Install Open3D

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

Install Open3D version 0.14.1 for visualization. Update pip first if necessary.

```bash
python3 -m pip install -U pip # run this if pip can not find this version
python3 -m pip install open3d==0.14.1

```

--------------------------------

### Install PyTorch with Conda (Default)

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Creates and activates a Conda environment named 'easymocap' with Python 3.9, then installs PyTorch and Torchvision using pip. This is a general installation command.

```bash
conda create -n easymocap python=3.9 -y
conda activate easymocap
python3 -m pip install torch torchvision

```

--------------------------------

### Install EasyMocap Core Components

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

This command sequence clones the EasyMocap repository, installs dependencies from `requirements.txt`, and develops the package in the current environment.

```bash
git clone https://github.com/zju3dv/EasyMocap.git
python3 -m pip install -r requirements.txt
python3 setup.py develop

```

--------------------------------

### Install Human Body Prior

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Run this script to install the human body prior module. Skip if not using this module.

```bash
bash ./scripts/install/install_vposer.sh
```

--------------------------------

### Start EasyAnnot Server

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/detect_chessboard.html

Start the EasyAnnot server and access it via the specified port in your browser. The root directory for data should be provided.

```bash
python3 app.py --root ${data}/ground1f --port 2345
```

--------------------------------

### Install detectron2

Source: https://chingswy.github.io/easymocap-public-doc/install/install_segment.html

Install detectron2 by specifying the CUDA and Torch versions. Ensure compatibility with your environment.

```bash
python -m pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu111/torch1.9/index.html
```

--------------------------------

### Install Instant-ngp

Source: https://chingswy.github.io/easymocap-public-doc/install/install_nvs.html

Clone the Instant-ngp repository and build it using CMake. Ensure you have a compatible environment.

```bash
mkdir -p 3rdparty && cd 3rdparty
git clone --recursive https://github.com/nvlabs/instant-ngp
cd instant-ngp
cmake . -B build
cmake --build build --config RelWithDebInfo -j 16

```

--------------------------------

### Install YOLOv5 and Download Weights

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/prepare_object.html

Clone the YOLOv5 repository and download pre-trained weights. Ensure you are in the correct directory before running these commands.

```bash
mkdir -p 3rdparty && cd 3rdparty
git clone https://github.com/chingswy/yolov5.git
wget https://github.com/ultralytics/yolov5/releases/download/v6.1/yolov5m.pt
mv yolov5m.pt models/

```

--------------------------------

### Install OpenPose on Linux

Source: https://chingswy.github.io/easymocap-public-doc/install/install_2d.html

Follow these commands to clone the OpenPose repository, update submodules, install dependencies, and build the project with Python support.

```bash
git clone https://github.com/CMU-Perceptual-Computing-Lab/openpose.git --depth 1
cd openpose
git submodule update --init --recursive --remote
sudo apt install libopencv-dev
sudo apt install protobuf-compiler libgoogle-glog-dev
sudo apt install libboost-all-dev libhdf5-dev libatlas-base-dev
mkdir build
cd build
cmake .. -DBUILD_PYTHON=true
make -j8

```

--------------------------------

### Install EasyMocap (2023.06.30 Update - CUDA 11.6)

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

This command installs EasyMocap with Python 3.9, CUDA 11.6, and PyTorch 1.12.0. It also installs `spconv-cu116`.

```bash
git clone https://github.com/zju3dv/EasyMocap.git
conda create -n easymocap python=3.9 -y
conda activate easymocap
wget -c https://download.pytorch.org/whl/cu116/torch-1.12.0%2Bcu116-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torch-1.12.0+cu116-cp39-cp39-linux_x86_64.whl
wget -c https://download.pytorch.org/whl/cu116/torchvision-0.13.0%2Bcu116-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torchvision-0.13.0+cu116-cp39-cp39-linux_x86_64.whl
python -m pip install -r requirements.txt
pip install spconv-cu116
# install pyrender if you have a screen
python3 -m pip install pyrender
python setup.py develop

```

--------------------------------

### Install easymocap Environment

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb.html

Install the necessary dependencies for the easymocap environment, including pytorch-lightning and spconv. Refer to 'requirements_neuralbody.txt' for a complete list.

```bash
pip install -r requirements_neuralbody.txt
```

--------------------------------

### Install NeuralBody Requirements

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb_install.html

Use this command to install all necessary Python packages for NeuralBody from the specified requirements file.

```bash
python3 -m pip install -r requirements_neuralbody.txt 

```

--------------------------------

### Install PyTorch3D (Conda)

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

Install PyTorch3D using Conda, including necessary dependencies like fvcore and iopath.

```bash
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub
conda install pytorch3d -c pytorch3d

```

--------------------------------

### Install Mediapipe

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to install the Mediapipe library using pip. This is a prerequisite for using the Mediapipe keypoint extraction mode.

```bash
python3 -m pip install mediapipe

```

--------------------------------

### Install SPConv for NeuralBody

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Install the spconv-cu111 package, a dependency for NeuralBody.

```python
python3 -m pip install spconv-cu111
```

--------------------------------

### Install SCHP Dependencies

Source: https://chingswy.github.io/easymocap-public-doc/install/install_segment.html

Clone the SCHP repository and install necessary Python packages using pip and conda.

```bash
cd 3rdparty
git clone https://github.com/chingswy/Self-Correction-Human-Parsing.git
cd Self-Correction-Human-Parsing
conda install ninja
python3 -m pip install scikit-image
python3 -m pip install networks
```

--------------------------------

### Install NeuralBody Requirements

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Install the necessary Python packages for the NeuralBody module using pip.

```python
python3 -m pip install -r requirements_neuralbody.txt
```

--------------------------------

### Install PyTorch for CUDA 10.1 / Python 3.7

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Installs PyTorch 1.7.1 and Torchvision 0.8.2 compiled for CUDA 10.1 and Python 3.7 using downloaded wheel files. Requires a Conda environment.

```bash
conda create -n easymocap python=3.7 -y
conda activate easymocap
wget -c https://download.pytorch.org/whl/cu101/torch-1.7.1%2Bcu101-cp37-cp37m-linux_x86_64.whl
wget -c https://download.pytorch.org/whl/cu101/torchvision-0.8.2%2Bcu101-cp37-cp37m-linux_x86_64.whl
python3 -m pip install ./torch-1.7.1+cu101-cp37-cp37m-linux_x86_64.whl
python3 -m pip install ./torchvision-0.8.2+cu101-cp37-cp37m-linux_x86_64.whl

```

--------------------------------

### Download Yolov4 Model Weight

Source: https://chingswy.github.io/easymocap-public-doc/install/install_2d.html

Download the yolov4 model weight and place it in the data/models directory. This is required for Yolov4+HRNet installation.

```bash
mkdir -p data/models
wget -c https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights
mv yolov4.weights data/models

```

--------------------------------

### Extract Keypoints with YOLOv4+Openpose (Cropped)

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to extract keypoints using YOLOv4 for human detection followed by OpenPose on cropped images. Requires OpenPose installation path.

```bash
python3 apps/preprocess/extract_keypoints.py ${data} --mode openposecrop --openpose ${openpose}

```

--------------------------------

### Copy and Sample Dataset

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/tools.html

This script copies a dataset and samples frames based on specified start, end, and step parameters. It can also resize images using a scale factor. Provide input and output directory paths.

```bash
python3 scripts/preprocess/copy_dataset.py /path/to/input /path/to/output

```

--------------------------------

### Install PyTorch for CUDA 10.0

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Installs PyTorch 1.4.0 and Torchvision 0.5.0 for CUDA 10.0 using wheel files. This command is specific to older CUDA versions.

```bash
python3 -m pip install torch-1.4.0+cu100-cp37-cp37m-linux_x86_64.whl
pip install https://download.pytorch.org/whl/cu100/torchvision-0.5.0%2Bcu100-cp37-cp37m-linux_x86_64.whl

```

--------------------------------

### Install PyTorch for CUDA 11.1 / Python 3.9

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Installs PyTorch 1.9.1 and Torchvision 0.10.1 specifically compiled for CUDA 11.1 and Python 3.9 using downloaded wheel files. Requires a Conda environment.

```bash
conda create -n easymocap python=3.9 -y
conda activate easymocap
wget -c https://download.pytorch.org/whl/cu111/torch-1.9.1%2Bcu111-cp39-cp39-linux_x86_64.whl
wget -c https://download.pytorch.org/whl/cu111/torchvision-0.10.1%2Bcu111-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torch-1.9.1+cu111-cp39-cp39-linux_x86_64.whl
python3 -m pip install ./torchvision-0.10.1+cu111-cp39-cp39-linux_x86_64.whl

```

--------------------------------

### Extract Keypoints with OpenPose (Body and Feet)

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to extract body and feet keypoints using OpenPose. Requires specifying the path to your OpenPose installation.

```bash
openpose=<path/to/openpose/installation>
# detect the body and feet keypoints
python3 apps/preprocess/extract_keypoints.py ${data} --mode openpose --openpose ${openpose}

```

--------------------------------

### Extract Keypoints with OpenPose (Hand and Face)

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to extract hand and face keypoints using OpenPose. This is in addition to body and feet keypoints and requires the OpenPose installation path.

```bash
# detect the hand and face if needed
python3 apps/preprocess/extract_keypoints.py ${data} --mode openpose --openpose ${openpose} --hand --face

```

--------------------------------

### Manually Clip Images for Annotation

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/capture_youtube.html

Initiates the manual clipping interface for annotating video segments. Use 'j' to mark the start and 'k' to mark the end of a clip, then 'l' to save it. Navigation is done with 'a/d' for frames and 'w/s' for 100-frame jumps.

```bash
python3 apps/annotation/annot_clip.py ${database}
```

--------------------------------

### Verify MANO Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Run this command to test the MANO model after placement.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/mano.yml
```

--------------------------------

### Download SPIN Models and Data

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/prepare_mono.html

Use these commands to download the pretrained SPIN model and additional data required for keypoint extraction. Ensure the target directories exist.

```bash
mkdir -p data/models
wget -O data/models/spin_checkpoints.pt http://visiondata.cis.upenn.edu/spin/model_checkpoint.pt
wget -c http://visiondata.cis.upenn.edu/spin/data.tar.gz
# data/models/smpl_mean_params.npz
```

--------------------------------

### Train and Visualize with Instant-NGP

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-dense.html

Train and visualize neural graphics primitives using Instant-NGP. This can be done interactively or by running a training script in the background.

```bash
# train and visualize ngp
cd 3rdparty/instant-ngp
data=${root}/human1f
# interactively
./build/testbed --scene ${root}/ngp-human1f
# or run in background
python3 scripts/run.py --scene ${data} --mode nerf --screenshot_transforms ${data}/transforms_novel.json --n_steps 100000 --width 1080 --height 1920 --screenshot_dir ${data}/output --save_snapshot ${data}/ckpt.msgpack

```

--------------------------------

### Download YouTube Video and Preprocess

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/prepare_mirror.html

Use this script to download a YouTube video and extract images, then annotate clips. The `--copy` flag creates a new folder for clipped annotations.

```bash
database=data/mirror-youtube
python3 scripts/dataset/download_youtube.py "https://www.youtube.com/watch?v=hVDPS-f6K5o" --database ${database}
python3 apps/preprocess/extract_image.py ${database}
python3 apps/annotation/annot_clip.py ${database}
python3 apps/annotation/annot_clip.py ${database} --copy --out data/mirror-youtube-clip
```

--------------------------------

### Run NeuralBody Demo for Rendering

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb.html

Execute the demo script to render novel views. The first command is recommended for 4x3090 GPUs. The second command is provided for lower-end GPUs but is not recommended.

```bash
# render with 4x3090
python3 apps/neuralbody/demo.py --mode soccer1_6 ${data} --gpus 0,1,2,3 --demo
```

```bash
# (not recommand) 
python3 apps/neuralbody/demo.py --mode soccer1_6 ${data} --gpus 0, data_share_args.sample_args.nrays 1024 --demo
```

--------------------------------

### Prepare SMPL Models Directory Structure

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

This illustrates the expected directory structure for SMPL models after downloading and extracting them to `data/bodymodels/`. It specifically shows the `smplv1.1.0` model.

```bash
data/bodymodels
└── SMPL_python_v.1.1.0
    └── smpl
        ├── __init__.py
        ├── models
        │   ├── basicmodel_f_lbs_10_207_0_v1.1.0.pkl
        │   ├── basicmodel_m_lbs_10_207_0_v1.1.0.pkl
        │   └── basicmodel_neutral_lbs_10_207_0_v1.1.0.pkl
        └── smpl_webuser

```

--------------------------------

### Write Camera Parameters using EasyMocap

Source: https://chingswy.github.io/easymocap-public-doc/database/1_camera.html

Shows how to write camera parameters to YAML files using the `write_cameras` utility function from `easymocap.mytools.camera_utils`. The `cameras` dictionary should map camera names to their respective parameter dictionaries, each containing 'K', 'dist', 'R', and 'T'.

```python
from easymocap.mytools.camera_utils import write_cameras
# cameras: dict
camera_00 = {
    'K': np.eye(3),
    'dist': np.zeros(5),
    'R': np.eye(3),
    'T': np.zeros(3),
}
camera_01 = {
    'K': np.eye(3),
    'dist': np.zeros(5),
    'R': np.eye(3),
    'T': np.zeros(3),
}
cameras = {
    '00': camera_00,
    '01': camera_01,
}
write_cameras(path, cameras)
```

--------------------------------

### Verify VPoser with SMPL+H Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Test the VPoser model with the SMPL+H model.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/smplh_vposer.yml
```

--------------------------------

### Run Hand Triangulation with EasyMocap

Source: https://chingswy.github.io/easymocap-public-doc/develop/01_triangulate.html

Execute the EasyMocap command-line tool to perform hand detection and triangulation. Ensure the dataset and configuration files are correctly specified.

```bash
emc --data config/datasets/mvimage.yml --exp config/mv1p/detect_hand_triangulate.yml --root ${data} --subs_vis 0 1 2
```

--------------------------------

### Verify VPoser with SMPL Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Test the VPoser model in conjunction with the SMPL model.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/smpl_vposer.yml
```

--------------------------------

### Read Camera Parameters using EasyMocap

Source: https://chingswy.github.io/easymocap-public-doc/database/1_camera.html

Demonstrates how to read camera intrinsic and extrinsic parameters from YAML files using the `read_cameras` utility function from `easymocap.mytools.camera_utils`. Ensure the path points to the directory containing `intri.yml` and `extri.yml`.

```python
from easymocap.mytools.camera_utils import read_cameras
# path/intri.yml, path/extri.yml
path = 'xxx' 
cameras = read_cameras(path)
```

--------------------------------

### SMPL Fitting Configuration Stages

Source: https://chingswy.github.io/easymocap-public-doc/develop/02_fitsmpl.html

Configuration snippet for the multi-stage SMPL fitting process. This outlines the loading of the SMPL model, initialization of parameters, and refinement of poses with various loss functions.

```yaml
...
  at_final:
    load_body_model: # Load SMPL model
      module: myeasymocap.io.model.SMPLLoader
      args:
        model_path: models/pare/data/body_models/smpl/SMPL_NEUTRAL.pkl #
        regressor_path: models/J_regressor_body25.npy
    init_params: ... # initialize the SMPL parameters
    fitShape: ... # Fit the SMPL shape by 3D limbs
    init_RT: ... # Initialize the body rotation and translation
    refine_poses: ... # Optimize the poses
      # ...
      args:
        optimize_keys: [poses, Rh, Th]
        loss:
          k3d: ... # 3D keypoints distance
          smooth: ... # smooth term
          prior: ... # use GMM as pose prior
    write: ... # write the results
    render_ground: ... # render the SMPL with a ground
    make_video: ...

```

--------------------------------

### Calibrate Dense Cameras with COLMAP

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-dense.html

Calibrate camera rigs using COLMAP, leveraging the human body for more feature points. Ensure COLMAP is installed and accessible.

```python
python3 apps/calibration/calib_dense_by_colmap.py ${root}/human1f ${root}/colmap-human1f --share_camera --colmap ${colmap}

```

--------------------------------

### Train NeuralBody Model

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb_code.html

Initiate the training process for the NeuralBody model. Specify the mode, GPUs to use, and the data path.

```bash
python3 apps/neuralbody/demo.py --mode soccer1_6 --gpus 0,1,2,3 ${data}
```

--------------------------------

### Extract Keypoints with YOLO-HRNet and OpenPose

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/prepare_mirror.html

Extracts keypoints using YOLO-HRNet. Optionally, use OpenPose for feet detection if installed. Tracks the human using a separate script.

```bash
python3 apps/preprocess/extract_keypoints.py ${data} --mode yolo-hrnet
# use OpenPose to detect the feet, skip it if you don't install OpenPose
python3 apps/preprocess/extract_keypoints.py ${data} --mode feetcrop --hand
# track the human
python3 apps/preprocess/extract_track.py ${data}
```

--------------------------------

### Preprocess SMPL Data for NeuralBody

Source: https://chingswy.github.io/easymocap-public-doc/develop/05_neuralbody.html

Convert SMPL parameters to vertices and render instance maps. Ensure the data path and output directories are correctly set.

```bash
data=<path/to/boxing>
mkdir -p ${data}/output-smpl-3d
cp -r output/boxing/smpl ${data}/output-smpl-3d
# convert the SMPL parameter to vertices
python3 apps/postprocess/write_vertices.py ${data}/output-smpl-3d/smpl ${data}/output-smpl-3d/vertices --cfg_model config/model/smpl.yml --mode vertices
# render the instance map
python3 apps/postprocess/render.py ${data} --exp output-smpl-3d --mode instance-d0.05 --ranges 0 200 1 --model config/model/smpl.yml
```

--------------------------------

### Extract Images from Video Data

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb.html

Use this script to extract individual image frames from compressed video files. Ensure the 'data' variable points to the correct path of your example data.

```bash
data=<path/to/example/data>
# extract the images
python3 apps/preprocess/extract_image.py ${data}
```

--------------------------------

### Detect Chessboard for Initialization

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Detects a chessboard in images for camera calibration initialization. Use `--check` for additional verification.

```bash
python3 apps/calibration/detect_chessboard.py ${root}/intri --out ${root}/intri/output --pattern 9,6 --grid 0.1
```

```bash
python3 apps/calibration/detect_chessboard.py ${root}/ground1f --out ${root}/ground1f/output --pattern 9,6 --grid 0.1 --check
```

--------------------------------

### OpenCV Camera Intrinsics YAML Format

Source: https://chingswy.github.io/easymocap-public-doc/database/1_camera.html

Example of how intrinsic camera parameters (K and dist) are stored in YAML format, compatible with OpenCV. This includes camera names and their corresponding matrices.

```yaml
names:
  - "01"
  - "03"
K_01: !!opencv-matrix
  rows: 3
  cols: 3
  dt: d
  data: [2272.816, 0.000, 1255.278, 0.000, 2269.930, 729.021, 0.000, 0.000, 1.000]
dist_01: !!opencv-matrix
  rows: 1
  cols: 5
  dt: d
  data: [0.057, -0.047, 0.002, -0.002, 0.000]
K_03: !!opencv-matrix
  rows: 3
  cols: 3
  dt: d
  data: [2272.816, 0.000, 1255.278, 0.000, 2269.930, 729.021, 0.000, 0.000, 1.000]
dist_03: !!opencv-matrix
  rows: 1
  cols: 5
  dt: d
  data: [0.057, -0.047, 0.002, -0.002, 0.000]
```

--------------------------------

### Set CUDA Environment Variables

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

Sets environment variables for CUDA path and library path. This is useful for ensuring the system can find the correct CUDA installation, especially when multiple versions are present.

```bash
# export CUDA_VER=10.0
# export CUDA_VER=10.1
export CUDA_VER=11.4
export PATH=/usr/local/cuda-$CUDA_VER/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-$CUDA_VER/lib64:$LD_LIBRARY_PATH

```

--------------------------------

### Initialize MatchAndTrack Class

Source: https://chingswy.github.io/easymocap-public-doc/develop/04_multiperson.html

Initializes the MatchAndTrack class with configuration for matching and tracking modules. The matcher handles multi-view matching, and the tracker manages individual tracking across frames.

```python
class MatchAndTrack():
    def __init__(self, cfg_match, cfg_track) -> None:
        self.matcher = MatchBase(cfg_match)
        self.tracker = TrackBase(cfg_track)

```

--------------------------------

### OpenCV Camera Extrinsics YAML Format

Source: https://chingswy.github.io/easymocap-public-doc/database/1_camera.html

Example of how extrinsic camera parameters (rotation vector R and translation vector T) are stored in YAML format, compatible with OpenCV. This includes rotation matrix Rot as well.

```yaml
names:
  - "01"
R_01: !!opencv-matrix
  rows: 3
  cols: 1
  dt: d
  data: [1.258, -1.292, 1.178]
Rot_01: !!opencv-matrix
  rows: 3
  cols: 3
  dt: d
  data: [-0.021, -1.000, -0.005, -0.087, 0.007, -0.996, 0.996, -0.021, -0.087]
T_01: !!opencv-matrix
  rows: 3
  cols: 1
  dt: d
  data: [0.453, 1.048, 4.957]
```

--------------------------------

### Launch COLMAP GUI

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Launches the COLMAP graphical user interface for database and image path management.

```bash
$colmap gui --database_path ${root}/calib/merged/database.db --image_path ${root}/calib/merged/images --import_path ${root}/calib/merged/sparse/0
```

--------------------------------

### Run Multi-Person Motion Capture

Source: https://chingswy.github.io/easymocap-public-doc/develop/04_multiperson.html

Execute the multi-person motion capture process using a specified configuration file and data root directory.

```bash
data=data/examples/ballet_2person
emc --cfg config/mvmp/body/ballet_2person.yml --root ${data}

```

--------------------------------

### Check Camera Calibration

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Verifies the camera calibration by generating and writing a cube model. Use `--show` to visualize.

```bash
python3 apps/calibration/check_calib.py ${root}/ground1f --out ${root}/ground1f --mode cube --write
```

--------------------------------

### Motion Capture with Monocular and Mirror Videos

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/quickstart.html

Execute motion capture using monocular and mirror videos. Adjust `fps` and `vis_scale` as needed.

```python
data=<path/to/data>
python3 apps/demo/mocap.py ${data} --work mirror --fps 30 --vis_scale 0.5

```

--------------------------------

### Visualize Camera Calibration with MV1P

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/debug_mv1p.html

These commands help visualize camera calibration results. Ensure extrinsic parameters are in meters. The 'cube' mode plots a unit cube at the origin, while the 'human' mode triangulates and projects human keypoints into each view.

```bash
# this command will plot a unit cube in the origin:
python3 apps/calibration/check_calib.py ${data} --out ${data} --mode cube --write --show
```

```bash
# this command will read and triangulate the keypoints of the human and project it to each views.
python3 apps/calibration/check_calib.py ${data} --out ${data} --mode human --write --show
```

--------------------------------

### Run Monocular Video Fitting Command

Source: https://chingswy.github.io/easymocap-public-doc/develop/03_fitsmpl_monocular.html

This command initiates the fitting process for monocular videos. Ensure the dataset and configuration files are correctly specified. The `--subs` argument selects the specific video to process.

```bash
data=data/examples/internet-rotate
emc --data config/datasets/svimage.yml --exp config/1v1p/hrnet_pare_finetune.yml --root ${data} --ranges 0 500 1 --subs 23EfsN7vEOA+003170+003670

```

--------------------------------

### Verify SMPL+H Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Execute this command to verify the SMPL+H model.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/smplh.yml
```

--------------------------------

### Extract Keypoints with YOLOv4+HRNet+Openpose (Feet)

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to extract keypoints using YOLOv4 for detection, HRNet for initial pose estimation, and OpenPose specifically for feet keypoints. The --force flag may be used to overwrite existing annotations.

```bash
python3 apps/preprocess/extract_keypoints.py ${data} --mode yolo-hrnet & python3 apps/preprocess/extract_keypoints.py ${data} --mode feetcrop --openpose ${openpose} --force

```

--------------------------------

### Verify FLAME Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Use this command to check the FLAME model.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/flame.yml
```

--------------------------------

### Train NeuralBody Model

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb.html

Initiate the training process for the NeuralBody model. Recommended for 4x3090 GPUs. For lower-end GPUs like RTX 1080Ti/3060, reduce the number of rays.

```bash
data=/path/to/dataset
# Recommand training with 4x3090
python3 apps/neuralbody/demo.py --mode soccer1_6 ${data} --gpus 0,1,2,3
```

```bash
# Reduce the number of rays if you train with RTX 1080Ti/3060
python3 apps/neuralbody/demo.py --mode soccer1_6 ${data} --gpus 0, data_share_args.sample_args.nrays 1024
```

--------------------------------

### Run Monocular Hand Fitting Command

Source: https://chingswy.github.io/easymocap-public-doc/develop/03_fitsmpl_monocular.html

This command is used to fit a MANO model to monocular hand videos. It requires specifying the dataset, experiment configuration, root directory, and the video to be processed using the `--subs` argument.

```bash
data=data/examples/hand
emc --data config/datasets/svimage.yml --exp config/1v1p/hand_detect_finetune.yml --root ${data} --ranges 0 1800 1 --subs video

```

--------------------------------

### Configure MediaPipe Hand Detection Module

Source: https://chingswy.github.io/easymocap-public-doc/develop/01_triangulate.html

Define the MediaPipe hand detection module within the EasyMocap configuration. Specify the module path, keypoint source, and the path to the MediaPipe hand landmark model.

```yaml
detect:
  module: myeasymocap.backbone.mediapipe.hand.MediaPipe
  key_from_data: [images, imgnames]
  args:
    ckpt: models/mediapipe/hand_landmarker.task
```

--------------------------------

### Set PYOPENGL_PLATFORM environment variable

Source: https://chingswy.github.io/easymocap-public-doc/install/install_vis3d.html

This command must be run once before using pyrender on a server.

```bash
export PYOPENGL_PLATFORM=osmesa

```

--------------------------------

### Check MV1P Datasets

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/debug_mv1p.html

Use these commands to verify the synchronization and accuracy of 2D keypoints within your dataset. The annotation tool helps in checking keypoints, and the dataset test script validates the overall dataset integrity.

```bash
# use the annotation tool to check the 2D keypoints
python3 apps/annotation/annot_keypoints.py ${data}
```

```bash
# check the dataset
python3 apps/fit/test_dataset.py --cfg_data config/data/mv1p.yml --opt_data args.path ${data} args.out ${data}/output-keypoints3d
```

--------------------------------

### Define Clip Rendering Parameters (Clip3)

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb_code.html

Configure rendering parameters for Clip3, including frame and view ranges, and object keys. This snippet demonstrates freezing motion and duplicating objects with transparency.

```yaml
frozen1:
  frame: [40, 200, 2]
  view: [200, 201, 1]
  object_keys: ["ground_@{'frame': 40}", 'ball_0', "human_3_@{'scale_occ': 0.2}", "human_0_@{'frame': 40}", "human_1_@{'frame': 40}", "human_2_@{'frame': 40}", "human_3_@{'frame': 40}", "human_4_@{'frame': 40}", "human_5_@{'frame': 40}"]
frozen2:
  frame: [199, 39, -2]
  view: [200, 201, 1]
  object_keys: ["ground_@{'frame': 40}", 'ball_0', "human_3_@{'scale_occ': 0.2}", "human_0_@{'frame': 40}", "human_1_@{'frame': 40}", "human_2_@{'frame': 40}", "human_3_@{'frame': 40}", "human_4_@{'frame': 40}", "human_5_@{'frame': 40}"]
```

--------------------------------

### Check NeuralBody Training Dataset

Source: https://chingswy.github.io/easymocap-public-doc/develop/05_neuralbody.html

Verify that the dataset can be loaded correctly for NeuralBody training. This script checks the training set configuration.

```python
python3 apps/neuralbody/check_trainset.py --cfg config/neuralbody/dataset/multiview/lightstage.yml --opts split train data_share_args.root ${data}
```

--------------------------------

### Extract Instance Segmentation

Source: https://chingswy.github.io/easymocap-public-doc/install/install_segment.html

Run the instance segmentation extraction script using your dataset and the path to your downloaded models. Consider using a temporary directory for large outputs.

```bash
python3 extract_multi.py ${data} --ckpt_dir /path/to/models
```

```bash
python3 extract_multi.py ${data} --ckpt_dir /path/to/models --tmp /path/to/tmp
```

--------------------------------

### Check MV1P SMPL Fitting

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/debug_mv1p.html

Run these commands to check the SMPL fitting process for different frame ranges. The `--exp` flag allows you to specify a debug output name.

```bash
# check only first 1 frames
python3 apps/demo/mocap.py ${data} --mode smpl-3d --ranges 0 1 1 --exp debug1
```

```bash
# check only first 10 frames
python3 apps/demo/mocap.py ${data} --mode smpl-3d --ranges 0 10 1 --exp debug10
```

```bash
# check only first 100 frames
python3 apps/demo/mocap.py ${data} --mode smpl-3d --ranges 0 100 1 --exp debug100
```

--------------------------------

### Copy Dataset Script

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-dense.html

Use this script to organize raw data into the required directory structure for calibration.

```python
python3 scripts/preprocess/copy_dataset.py <path/to/rawdata> ${root}/ground1f --start 0 --end 1

```

--------------------------------

### Check Calibration Results

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration

Verifies the calibration results by checking against a cube model. Options include showing the results and writing them to a file. Useful for validating extrinsic calibration.

```bash
python3 apps/calibration/check_calib.py ${ground} --mode cube --out ${ground} --show --write
```

--------------------------------

### Verify SMPL Model

Source: https://chingswy.github.io/easymocap-public-doc/install/install_smpl.html

Use this command to check if the SMPL model has been correctly placed and is accessible.

```bash
python3 apps/vis3d/vis_smpl.py --cfg config/model/smpl.yml
```

--------------------------------

### Render Scene with NeuralBody

Source: https://chingswy.github.io/easymocap-public-doc/works/multinb_code.html

Execute the rendering process for a scene using the NeuralBody model. Specify the mode, GPUs, and demo data path.

```bash
python3 apps/neuralbody/demo.py--mode soccer1_6 --gpus 0,1,2,3 --demo ${data}
```

--------------------------------

### Read and Write Camera Parameters

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration

Use these utility functions to read camera intrinsic and extrinsic parameters from a file and write them to a specified directory. Ensure the 'T' metric is in meters.

```python
from easymocap.mytools.camera_utils import read_cameras, write_camera

# path/intri.yml, path/extri.yml
path = 'xxx' 
cameras = read_cameras(path)
write_camera(outdir, cameras)
```

--------------------------------

### Full Motion Capture and Data Preparation

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/quickstart.html

A comprehensive command for motion capture and data preparation, including options to disable cropping.

```python
# motion capture
python3 apps/demo/mocap.py ${data} --work lightstage-dense-unsync --subs_vis 01 07 13 19 --disable_crop

```

--------------------------------

### Annotate Chessboard for Calibration

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration

Use this script to annotate detected chessboards, which is a necessary step before extrinsic calibration. Ensure the annotation mode is set to 'chessboard'.

```bash
python3 apps/annotation/annot_calib.py ${data} --annot chessboard --mode chessboard
```

--------------------------------

### Calibrate Extrinsic Parameters

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration

Calibrates the extrinsic parameters of the cameras using detected chessboard information. Requires ground truth data and intrinsic parameters.

```bash
ground=/nas/datasets/ZJUMoCap-v1/ground
ba=/nas/datasets/ZJUMoCap-v1/ba
python3 apps/calibration/calib_extri.py ${ground} --intri ${ba}/output/intri.yml
```

--------------------------------

### Calibrate Sparse Scene with COLMAP

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Initializes and calibrates a sparse scene using COLMAP, referencing ground truth data for initialization.

```bash
python3 apps/calibration/calib_sparse_by_colmap.py ${root}/human519 --init ${root}/ground1f --out /mnt/data2/shuai/calib-zjumocap --colmap ${colmap}
```

--------------------------------

### Extract Keypoints with Mediapipe Holistic

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/keypoints.html

Command to run keypoint detection using the Mediapipe Holistic model, which detects full body, hands, and face.

```bash
python3 apps/preprocess/extract_keypoints.py ${data} --mode mp-holistic

```

--------------------------------

### Dataset Configuration (MVDataset)

Source: https://chingswy.github.io/easymocap-public-doc/develop/01_triangulate.html

Configure the dataset parameters using a YAML file. This specifies the root directory, visualized views, and image reading settings.

```yaml
module: myeasymocap.datasets.mv1p.MVDataset
args:
  root: TO_BE_FILLED
  subs: []  # used views, default all views
  subs_vis: ['01'] # visualized views
  ranges: [0, 10000, 1]
  read_image: True
  reader:
    images:
      root: images
      ext: .jpg
    image_shape:
      root: images
      ext: .jpg
    cameras:
      root: ''

```

--------------------------------

### Run Fixed Hand Pose Fitting Command

Source: https://chingswy.github.io/easymocap-public-doc/develop/03_fitsmpl_monocular.html

This command executes the fitting process for scenarios with fixed hand poses, often used when estimating object poses from hand grasps. It utilizes a specific configuration file (`fixhand.yml`) designed for this purpose.

```bash
data=data/examples/3_Giuliano
emc --data config/datasets/svimage.yml --exp config/1v1p/fixhand.yml --root ${data} --ranges 0 1800 1

```

--------------------------------

### Clone EasyAnnot Repository

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/detect_chessboard.html

Clone the EasyAnnot repository to use the Web Annotation Tool.

```bash
git clone https://github.com/chingswy/EasyAnnot.git
```

--------------------------------

### Visualize COLMAP Results

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-dense.html

Visualize the calibration results generated by COLMAP using the COLMAP GUI. This requires the database, image path, and sparse reconstruction path.

```bash
$colmap gui --database_path ${root}/colmap-human1f/database.db --image_path ${root}/colmap-human1f/images --import_path ${root}/colmap-human1f/sparse/0

```

--------------------------------

### Run COLMAP for Sparse Video Calibration

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Processes sparse video data using COLMAP for calibration, sharing camera parameters and processing in steps.

```bash
python3 apps/calibration/calib_pycolmap.py ${root} ${root}/calib --share_camera --step 100 --seqs 506 508
```

--------------------------------

### Optional Body Model Directory Structure

Source: https://chingswy.github.io/easymocap-public-doc/install/install.html

This shows the directory structure for various optional body models, including FLAME2020, manov1.2, smplhv1.2, SMPL_python_v.1.1.0, and vposer_v02.

```bash
./data/bodymodels
├── FLAME2020
│   ├── flame_dynamic_embedding.npy
│   ├── FLAME_FEMALE.pkl
│   ├── FLAME_MALE.pkl
│   ├── FLAME_NEUTRAL.pkl
│   ├── flame_static_embedding.pkl
│   └── Readme.pdf
├── manov1.2
│   ├── MANO_LEFT.pkl
│   └── MANO_RIGHT.pkl
├── smplhv1.2
│   ├── female
│   ├── info.txt
│   ├── LICENSE.txt
│   ├── male
│   └── neutral
├── SMPL_python_v.1.1.0
│   └── smpl
└── vposer_v02
    ├── snapshots
    ├── V02_05.log
    └── V02_05.yaml

```

--------------------------------

### Annotate Calibration Data

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/studio-sparse.html

Manually annotates calibration data using a chessboard if automatic detection fails.

```bash
python3 apps/annotation/annot_calib.py ${root}/ground1f --mode chessboard --annot chessboard
```

--------------------------------

### Multi-Stage Experimental Configuration

Source: https://chingswy.github.io/easymocap-public-doc/develop/01_triangulate.html

Define the experimental stages for processing video data, including per-frame operations and finalization steps like smoothing and writing results.

```yaml
# The 'module' declares the multi-stage class used
module: myeasymocap.stages.basestage.MultiStage
args:
  output: output/detect_triangulate
  at_step: # This declares the operations to be performed on each frame of data
    detect: ... # detect the human by YOLO
    vis_det: ... # visualize the detection results
    keypoints2d: ... # estimate the 2D keypoints by HRNet
    vis2d: ... # visualize the human keypoints
    triangulate: ... # In this task, we only perform triangulation
    visualize: ... # Visualize the reprojection of each frame
  at_final: # This declares the operations that will be performed on all frames after completing the operations on each frame
    smooth: ... # Here we perform the smoothing of the skeleton
    write: ... # write the 3D keypoints
    make_video: ...

```

--------------------------------

### Create Markers Manually

Source: https://chingswy.github.io/easymocap-public-doc/quickstart/calibration/detect_chessboard.html

Use this command to create custom markers for manual annotation. Specify the output directory, number of markers per group, and number of groups.

```bash
python3 apps/calibration/create_marker.py ${root}/background1f --N 4 --N_group 4
```