This project implements the functions from pytorch360convert as custom ComfyUI nodes, along with a few additional useful functions for working with equirectangular 360 images and cubemaps.
The pytorch360convert library provides powerful and differentiable image transformation utilities for converting between different panoramic image formats like Equirectangular (360°) Images, Cubemap Representations, and Perspective Projections.
| E2C -> C2E | E2C -> Split -> Rebuild -> C2E |
|---|---|
 |
 |
If the provided installation information below is not enough, then please see here for more details on installing custom ComfyUI nodes: https://www.comflowy.com/advanced/how-to-install-comfyui-extension
If ComfyUI-Manager is installed, then you can download via pasting in the github link to this repo, or via the built-in install menu.
Download this repo as a zip file and then unzip it inside the ComfyUI custom_nodes directory.
Then install the dependency:
python -m pip install pytorch360convert
If using Windows, you can install git from here: https://git-scm.com/download/win if you have not done so already and wish to use Git. Then navigate to the custom nodes directory and then run:
git clone https://github.com/ProGamerGov/ComfyUI_pytorch360convert --recursive
Then install the dependency for Linux and MacOS:
python -m pip install pytorch360convert
Or for Windows:
path/to/python.exe -m pip install pytorch360convert
comfy node registry-install comfyui-pytorch360convert
See the provided workflow files in the examples directory for examples of how to use each new node.
The custom nodes added by this project are available under the 'pytorch360convert' heading in the node selection tab.
The E2C node converts equirectangular images into cubemaps. For more information see the pytorch360convert repo here.
The C2E node converts cubemaps into equirectangular images. For more information see the pytorch360convert repo here.
The E2P node converts equirectangular images into perspective images. For more information see the pytorch360convert repo here.
The E2E node rotates an equirectangular image along one or more axes (roll, pitch, and yaw) to produce a horizontal shift, vertical shift, or to roll the image. For more information see the pytorch360convert repo here.
The Roll Image node rolls an image along the height dimension (y axis) or the width dimension (x axis). This can help make artifacts like seams more accessible. Note that y axis rotations are not correct for final equirectangular outputs. Use the E2E node instead for y axis rotations.
These two nodes allow you to crop a section of an image out to make things like in-painting faster and more memory efficient, before pasting the cropped section back into the original image.
By default, the cropping node center crops images according to the provided crop_h and crop_w values. If crop_h2 and crop_w2 are used, then the image will be cropped according to [..., crop_h:crop_h2, crop_w:crop_w2, ...].
This is an experimental node for only pasting the modified parts of face images to the original equirectangular image. This is meant to help lessen the loss of image quality caused by converting between equirectangular and cubemap formats.
For viewing 360 images inside ComfyUI, see the github.com/ProGamerGov/ComfyUI_preview360panorama custom node.
Contributions are welcome! Please feel free to submit a Pull Request.
If you use this library in your research or project, please refer to the included CITATION.cff file or cite it as follows:
@misc{egan2024pytorch360convert,
title={PyTorch 360° Image Conversion Toolkit},
author={Egan, Ben},
year={2024},
publisher={GitHub},
howpublished={\url{https://github.com/ProGamerGov/pytorch360convert}}
}Egan, B. (2024). PyTorch 360° Image Conversion Toolkit [Computer software]. GitHub. https://github.com/ProGamerGov/pytorch360convert