(IET Lucknow'27) Freshman code push - Aditya 🐱👤
The official implementation of "Recurrent Person Image Generation for Pose Transfer, Virtual Try-on and Outfit Editing." by Aditya,
🔔 Updates
- [2023/10] Offical Colab Demo is now available at
. Data downloading and environment installation are included.
- [2023/10] Please check our latest version of paper for the updated and clarified implementation details.
- Clarification: the facial component was not added to the skin encoding as stated in the Techfest-:IITB 2023 workshop paper due to a minor typo. However, this doesn't affect our conclusions nor the comparison with the prior work, because it is an independent skin encoding design.
- [2023/20] To appear in Techfest IITB 2023.
- [2023/10] The best paper at Computer Vision for Fashion, Art and Design Workshop CVPR 2021.
Supported Try-on Applications
Supported Editing Applications
More results
Play with 
A directly runable demo can be found in our Colab!
Deepfashion Dataset can be found from DeepFashion MultiModal Source.
To set up the dataset in your specified data folder $DATA_ROOT, run:
pip install --upgrade gdown
python tools/download_deepfashion_from_google_drive.py --dataroot $DATA_ROOT
This script will automatically download all the necessary data from Google Drives (
images source, parse source, annotation source) to your the specified $DATA_ROOT in desired format.
Please install the environment based on your need.
Required packages are
torch
torchvision
tensorboardX
scikit-image==0.16.2
The version of torch/torchvison is not restricted for inference.
Note the training process requires CUDA functions provided by GFLA, which can only compile with torch=1.0.0.
To start training, please follow the installation instruction in GFLA to install the environment.
Then run pip install -r requirements.txt.
The pretrained weights can be found here. Please unzip them under checkpoints/ directory.
(The checkpoints above are reproduced, so there could be slightly difference in quantitative evaluation from the reported results. To get the original results, please check our released generated images here.)
(DIORv1_64 was trained with a minor difference in code, but it may give better visual results in some applications. If one wants to try it, specify --netG diorv1.)
Warmup the Global Flow Field Estimator
Note, if you don't want to warmup the Global Flow Field Estimator, you can extract its weights from GFLA by downloading the pretrained weights GFLA from here. (Check Issue #23 for how to extract weights from GFLA.)
Otherwise, run
sh scripts/run_pose.sh
Training
After warming up the flownet, train the pipeline by
sh scripts/run_train.sh
Run tensorboard --logdir checkpoints/$EXP_NAME/train to check tensorboard.
Note: Resetting discriminators may help training when it stucks at local minimals.
Download Generated Images
Here are our generated images which are used for the evaluation reported in the paper. (Deepfashion Dataset)
SSIM, FID and LPIPS
To run evaluation (SSIM, FID and LPIPS) on pose transfer task:
sh scripts/run_eval.sh
please always specific --frozen_flownet for inference.
If you find this work is helpful, please consider starring 🌟 this repo and citing us as
@InProceedings{Aditya_2023_Techfest,
author = {Aditya,Vineet},
title = {Recurrent Person Image Generation for Pose Transfer, Virtual Try-On and Outfit Editing},
month = {October},
year = {2023},
}
This repository is built up on GFLA, pytorch-CycleGAN-and-pix2pix, PATN and MUNIT. Please be aware of their licenses when using the code.
Thanks a lot for the great work to the pioneer researchers!