Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/ofnote/tryongan

TryOnGAN: Unofficial Implementation
https://github.com/ofnote/tryongan

deep-learning fashion gan generative-adversarial-network pytorch viton

Last synced: about 15 hours ago
JSON representation

TryOnGAN: Unofficial Implementation

Host: GitHub
URL: https://github.com/ofnote/tryongan
Owner: ofnote
License: other
Created: 2021-08-12T09:35:51.000Z (over 3 years ago)
Default Branch: main
Last Pushed: 2021-09-10T14:17:40.000Z (about 3 years ago)
Last Synced: 2024-05-14T00:15:22.907Z (6 months ago)
Topics: deep-learning, fashion, gan, generative-adversarial-network, pytorch, viton
Language: Python
Homepage:
Size: 2.61 MB
Stars: 85
Watchers: 5
Forks: 16
Open Issues: 5
Metadata Files:
- Readme: README.md
- License: LICENSE.txt

Awesome Lists containing this project

README

        # TryOnGAN-pytorch

WIP : Unofficial pytorch implementation of [TryOnGAN](https://arxiv.org/pdf/2101.02285.pdf)

There are multiple branches :

* `UC` : Unconditioned StyleGAN2-ada, without any changes.

* `PC-add` : Pose conditioned with pose encoders outputs added to style block outputs.

* `PC-concat` : Pose conditioned with pose encoder's outputs concatenated to style block outputs.

## Useful Notebooks :

* Alphapose notebook to create pose keypoints [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1vGjUZ8rm5HdkAOzqaM3cfYXUH3WaHH8L?usp=sharing)

* Training Notebook for UC : [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://github.com/confifu/TryOnGAN-pytorch/blob/main/notebooks/UC.ipynb)

* Training Notebook for PC : [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://github.com/confifu/TryOnGAN-pytorch/blob/main/notebooks/PC.ipynb)

## Useful links (Drive) :

* [Trained Checkpoints](https://drive.google.com/drive/folders/1_GxdMeMDhGcFlGowQg10dANO1liS0v4Z?usp=sharing)

* [Image Dataset](https://drive.google.com/file/d/1r_Pz2gk1Sp2sjYS7cj9oYUMbLGo9EyZS/view?usp=sharing)

* [Pose keypoints](https://drive.google.com/file/d/1XryABeNxvzsdjEY7j4ed2m4G-we0n0md/view?usp=sharing)

## Some interesting results:

* Interpolation

Color                      |  Length                  | Region

:-------------------------:|:-------------------------:|:-------------------------:

![](images/color_interpolation.png) |![](images/length_interpolation.png)|![](images/region_interpolation.png)

I inverted two real images into latent codes and interpolated between the two codes. Interestingly the intermediate images are valid garments. Maybe we can extend this technique to generate novel garments composed of multiple components. Note that pose changes for UC models but not with PC. Scratch here means model was trained with random weights without transfer learning.

* Style mixing for UC model

 

A is images generated by mixing latents of a real image with random latent code. B is by mixing two real images. Mixing here means replacing latent codes for lower six style blocks with other latent codes.

## Datasets

Some of these datasets have images only of dress or have multiple people in the same image. Such images need to be discarded or modified.

* [Deep Fashion(48k)](https://drive.google.com/drive/folders/0B7EVK8r0v71pTHhMenkxbE9fTVk?resourcekey=0-vTjoKh5LuocS_K3yxycXIg&usp=sharing)

* [Categorization Dress Pattern (16k)](https://data.world/crowdflower/categorization-dress-patterns)

* [Female Model images(2k)](https://www.kaggle.com/mercymeilya/2243-pictures-of-female-models-from-modelmangemt)

* [Apparel Dataset(16k)](https://www.kaggle.com/kaiska/apparel-dataset)

* [Bridal Dress(622)](https://www.kaggle.com/elhamazizi/bridal-dress)

* [Dress Recognition(7k)](https://www.kaggle.com/guardian222/dress-recognition)

* [Dress Fashion(60k)](https://www.kaggle.com/ssr123456/dressfasion)

* [Dress Class Color(8k)](https://www.kaggle.com/ssr123456/dress-class-color)

* [Trousers(1.4k)](https://www.kaggle.com/ssr123456/dress-trousers)

* [Apparel Images Dataset(11k)](https://www.kaggle.com/trolukovich/apparel-images-dataset)

* [Full-body-mads-dataset(1.2k)](https://www.kaggle.com/tapakah68/segmentation-full-body-mads-dataset)

* [Full-body-tiktok-dancing-dataset(2.6k)](https://www.kaggle.com/tapakah68/segmentation-full-body-tiktok-dancing-dataset)

* [somaset- synthetic humans(100k)](https://www.kaggle.com/vicolab/somaset)

* [Agender, full body image of people in the wild(4.6k)](https://www.kaggle.com/mayank735/agender)

* [Everybody dance now, single person,full body and pose keypoints(83.9k)](https://www.kaggle.com/amoghmisra27/everybody-dance-now1)

* [Everybody dance now(40.3k)](https://www.kaggle.com/lalitmohan4398/everybody-dance-nowsubject-2)

* [Everybody dance now(35.4k](https://www.kaggle.com/lalitmohan4398/everbody-dance-now-subject-3)

* [Yoga Poses(1.5k)](https://www.kaggle.com/niharika41298/yoga-poses-dataset)

* [Yoga Poses - Large(6k)](https://www.kaggle.com/shrutisaxena/yoga-pose-image-classification-dataset)

* [Human 3.6m](http://vision.imar.ro/human3.6m/description.php)

StyleGAN2 code is based on official [StyleGAN2-ada rep](https://github.com/NVlabs/stylegan2-ada-pytorch)