easy_dwpose

Recently, I tried to inference the DWPose (improved OpenPose) preprocessor for Diffusers and was shocked by how complicated it actually is! So, I decided to change that!

The goal of Easy DWPose is to provide a generic, reliable, and easy-to-use interface for making skeletons for ControlNet.

Me:

Why you should use it 😋

1. Easy installation!
2. Automatic checkpoint downloading.
3. Generic class to either import in Jupyter or to run on a video/images.
4. Code that is easy to read and modify.
5. Choose GPU for multi-gpu inference!
6. Custom drawing functions: convenient interface for modifying how you draw skeletons.

Installation

PIP

pip install easy-dwpose

From source

git clone git@github.com:reallyigor/easy_dwpose.git
cd easy_dwpose
pip install -e .

Quickstart

In you own .py scrip or in Jupyter

import torch
from PIL import Image

from easy_dwpose import DWposeDetector

# You can use a different GPU, e.g. "cuda:1"
device = "cuda:0" if torch.cuda.is_available() else "cpu"
detector = DWposeDetector(device=device)
input_image = Image.open("assets/pose.png").convert("RGB")

skeleton = detector(input_image, output_type="pil", include_hands=True, include_face=True)
skeleton.save("skeleton.png")

Input	Output

On a video

python scripts/inference_on_video.py --input assets/dance.mp4 --output_path result.mp4

Input	Output

On a folder of images

python scripts/inference_on_folder.py --input assets/ --output_path results/

Custom skeleton drawing

By default, we use standart skeleton drawing function but several projects change it (e.g. MusePose). Modify it or write your own from scratch!

from PIL import Image

from easy_dwpose import DWposeDetector
from easy_dwpose.draw.musepose import draw_pose as draw_pose_musepose

detector = DWposeDetector(device="cpu")
input_image = Image.open("assets/pose.png").convert("RGB")

skeleton = detector(input_image, output_type="pil", draw_pose=draw_pose_musepose, draw_face=False)
skeleton.save("skeleton.png")

Acknowledgement

We thank the original authors of the DWPose for their incredible models!

Thanks for open-sourcing!