This repository presents a Python-based tool designed to extract and visualize human pose keypoints utilizing OpenPose models within the ComfyUI environment. The OpenPoseNode class enables users to input single images to obtain both the keypoints and limbs, which can be displayed with customizable transparency settings.
- Allows downloading of OpenPose models directly from the Hugging Face Hub, saving them to a specified directory within ComfyUI.
- Processes individual images to extract human pose keypoints, offering visual outputs with adjustable transparency and a black background option for keypoints.
- Generates a JSON file containing the keypoints data for further analysis or integration.
Context
This tool serves as an extension for ComfyUI, specifically aimed at enhancing image processing capabilities by leveraging OpenPose models to extract and visualize human poses. Its primary function is to facilitate the analysis of human movements in images by providing both visual and data outputs.
Key Features & Benefits
Key functionalities include the ability to download OpenPose models seamlessly, which ensures that users can access the latest model versions without manual intervention. The tool also supports the extraction of pose keypoints from input images, offering visual feedback by overlaying keypoints and limbs directly on the images. Furthermore, it provides a JSON output of keypoints, making it easier for users to utilize this data in various applications.
Advanced Functionalities
The OpenPoseNode supports two distinct representations of human pose data: COCO and BODY_25. This flexibility allows users to select the model that best fits their specific requirements or preferences, enhancing the adaptability of the tool for different use cases.
Practical Benefits
This tool significantly streamlines workflows in ComfyUI by automating the process of pose extraction and visualization, thus saving time and improving control over image analysis tasks. By providing both visual outputs and structured data formats, users can enhance the quality of their projects and facilitate more detailed analyses of human poses.
Credits/Acknowledgments
The development of this tool was inspired by the work of beingjoey/pytorch_openpose_body_25, whose contributions laid the groundwork for this custom node. This project is licensed under the GPL-3.0 License, with further details available in the LICENSE file.