surajpatra700 / cp-vton-plus-3

CP-VTON+ (new 2023)

Original Author

@InProceedings{Minar_CPP_2020_CVPR_Workshops,
	title={CP-VTON+: Clothing Shape and Texture Preserving Image-Based Virtual Try-On},
	author={Minar, Matiur Rahman and Thai Thanh Tuan and Ahn, Heejune and Rosin, Paul and Lai, Yu-Kun},
	booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
	month = {June},
	year = {2020}
}

Project page
Saved/Pre-trained models: Checkpoints
Dataset: VITON_PLUS

what's new [Oct 2023]

• I upgrade the project to new version of libraries.
• Python 3.11 support.
  • torch=2.0.1+cu118
  • torchvision=0.15.2+cu118
  • opencv = 4.8.1.78
• Run app.py for testing or training.
• it can automatically run both commands (GMM and TOM) and take care of copying files.
• For training / testing
  • subprocess.call(gmm_train/gmm_test, shell=True)
  • subprocess.call(tom_train/tom_test, shell=True)
• fix all the deprecated warning of torch and resolve all isuses regarding dependency.
• have a dedicated branch for only-cpu version.

if you find any problem feel free to raise issue.

Installation and Run

1. create and virtual env.
2. if you are running on cpu, then follow this branch. CPU
3. if you have cuda then install torch with cuda. refer torch

after that, install the dependencies.

pip install -r requirements.txt

AutoRun

Run python app.py

for tensorboard Run tensorboard --logdir tensorboard

Training

https://github.com/minar09/cp-vton-plus#training

Testing

https://github.com/minar09/cp-vton-plus#testing

Pre-trained Models and datasets

Create checkpoints folder and copy the models to checkpoints/ Checkpoints
Create data folder copy the datasets to data/ VITON_PLUS

Testing with custom images

Refer my other repo for this, vtryon-app