Is that DensePose must be installed for deploy in ComfyUI？
Closed this issue · 11 comments
OS:win10
python：3.10
torch:2.1.2
cuda:12.1
when i deploy this node in comyui, it has come below import error,it seems like DensePose is missing. so does that DensePose have to be installed？
Yes, since Gradio and ComfyUI App involve the functionality of automatically generating Masks, DensePose is required to guide the process, hence the need to additionally install Detectron2 and DensePose. We have updated the description in the installation guide. If you only need to perform inference on a dataset, these two packages are not required.
i found that its very hard to install DensePose in Windows OS, could you provide a DensePose wheel file for Windows OS user？
DensePose is maintained by Facebook's Detectron, and we do not have experimental machines in the Windows environment, making it difficult for us to provide specific Windows installation instructions. However, I have found similar issues encountered by others, and I hope they can be of help to you, refer to this GitHub issue: vid2densepose #14.
thanks for your guideline，I have succeed import catVTON node in Comfyui. So if I need to download any dataset or pre-trained weight for this project now？thx~
For ease of use, we have integrated the weight download into the code, and the weights will be automatically downloaded to the local area the first time you run the workflow.
Of course, you can also manually download the weights from HuggingFace and fill in the local paths for sd15_inpainting and catvton into the "Model Loading" text box.
For datasets, if you do not need to evaluate on public datasets (which is usually required for scientific research) and are only using CatVTON for testing and playing, you can directly upload any suitable images without considering the download of the datasets.
It looks like the clothing change mission failed in two different types of dressup missions
The Chinese localization version of ComfyUI has overlaid the original input image Titles：
The input image on the left should be the person；
The input image on the right should be the reference garment / person wearing garments you want to transfer.
And mask sure the cloth_type in Auto Mask Generation match the garment you want to try on.
refer to the following image:
thx for your work, it seems like test success now!
😁 Have fun with CatVTON !
This issue is closed and if there is any other question in Windows OS, please come to #8 .