GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → bryandlee → naver-webtoon-faces

bryandlee / naver-webtoon-faces

Licence: other
Generative models on NAVER Webtoon faces

Programming Languages

Jupyter Notebook
11667 projects
python
139335 projects - #7 most used programming language

Labels

image-generationimage-to-image-translationface2cartoon

Projects that are alternatives of or similar to naver-webtoon-faces

SuperStyleNet
SuperStyleNet: Deep Image Synthesis with Superpixel Based Style Encoder (BMVC 2021)
Stars: ✭ 28 (-88.98%)
Mutual labels:  image-generation, image-to-image-translation
Awesome-ICCV2021-Low-Level-Vision
A Collection of Papers and Codes for ICCV2021 Low Level Vision and Image Generation
Stars: ✭ 163 (-35.83%)
Mutual labels:  image-generation, image-to-image-translation
Pix2pix
Image-to-image translation with conditional adversarial nets
Stars: ✭ 8,765 (+3350.79%)
Mutual labels:  image-generation, image-to-image-translation
AsymmetricGAN
[ACCV 2018 Oral] Dual Generator Generative Adversarial Networks for Multi-Domain Image-to-Image Translation
Stars: ✭ 42 (-83.46%)
Mutual labels:  image-generation, image-to-image-translation
OASIS
Official implementation of the paper "You Only Need Adversarial Supervision for Semantic Image Synthesis" (ICLR 2021)
Stars: ✭ 232 (-8.66%)
Mutual labels:  image-generation, image-to-image-translation
sText2Image
Attribute-Guided Sketch-to-Image Generation
Stars: ✭ 67 (-73.62%)
Mutual labels:  image-generation
UPIT
A fastai/PyTorch package for unpaired image-to-image translation.
Stars: ✭ 94 (-62.99%)
Mutual labels:  image-to-image-translation
PanoDR
Code and models for "PanoDR: Spherical Panorama Diminished Reality for Indoor Scenes" presented at the OmniCV workshop of CVPR21.
Stars: ✭ 22 (-91.34%)
Mutual labels:  image-to-image-translation
TimeSeries2DBarChartImageCNN
Conversion of the time series values to 2-D stock bar chart images and prediction using CNN (using Keras-Tensorflow)
Stars: ✭ 25 (-90.16%)
Mutual labels:  image-generation
NeuralTexGen
Image-space texture optimization of 3D meshes using PyTorch
Stars: ✭ 60 (-76.38%)
Mutual labels:  image-generation
texturize
🤖🖌️ Generate photo-realistic textures based on source images. Remix, remake, mashup! Useful if you want to create variations on a theme or elaborate on an existing texture.
Stars: ✭ 495 (+94.88%)
Mutual labels:  image-generation
image generator 2
Progressive GANによる画像生成器
Stars: ✭ 31 (-87.8%)
Mutual labels:  image-generation
SDEdit
PyTorch implementation for SDEdit: Image Synthesis and Editing with Stochastic Differential Equations
Stars: ✭ 394 (+55.12%)
Mutual labels:  image-generation
DeepSIM
Official PyTorch implementation of the paper: "DeepSIM: Image Shape Manipulation from a Single Augmented Training Sample" (ICCV 2021 Oral)
Stars: ✭ 389 (+53.15%)
Mutual labels:  image-to-image-translation
nemar
[CVPR2020] Unsupervised Multi-Modal Image Registration via Geometry Preserving Image-to-Image Translation
Stars: ✭ 120 (-52.76%)
Mutual labels:  image-to-image-translation
netflix-show-generator
A tool for generating Netflix show image
Stars: ✭ 18 (-92.91%)
Mutual labels:  image-generation
art2real
Art2Real: Unfolding the Reality of Artworks via Semantically-Aware Image-to-Image Translation. CVPR 2019
Stars: ✭ 71 (-72.05%)
Mutual labels:  image-to-image-translation
overlord
Official pytorch implementation of "Scaling-up Disentanglement for Image Translation", ICCV 2021.
Stars: ✭ 35 (-86.22%)
Mutual labels:  image-to-image-translation
Continuous-Image-Autoencoder
Deep learning image autoencoder that not depends on image resolution
Stars: ✭ 20 (-92.13%)
Mutual labels:  image-generation
ganslate
Simple and extensible GAN image-to-image translation framework. Supports natural and medical images.
Stars: ✭ 17 (-93.31%)
Mutual labels:  image-to-image-translation
View All Similar Projects ➔

Naver Webtoon Faces

./imgs/face2webtoon/dh.gif

Face cartoonization code is added. See here.

./imgs/face2webtoon/grb.jpg
./imgs/face2webtoon/je.jpg
./imgs/face2webtoon/jyp.jpg

This repo heavily uses codes from [this repo]

Dataset [link]

256*256 cartoon face images collected from on-going NAVER Webtoon series.

num titles 58
total images: 17662

./imgs/dataset.png

StyleGAN2

[paper] [model]

Training detail: batch 12, transfer from FFHQ, non-leaking augmentation

FID (10k samples)

Iteration FID
FFHQ 256.54
50k 11.73
100k 9.29
150k 8.87
200k 8.11
250k 7.41

Samples (FID 8.87)

./imgs/stylegan2/samples.png

4-Way Linear interpolation in w-space

./imgs/stylegan2/1.png

./imgs/stylegan2/2.png

Swapping codes at different layers

./imgs/stylegan2/3.png

Swapping Autoencoder for Deep Image Manipulation

[paper] [model]

SwapAE is a fully unsupervised generative model that embeds images into structure and style codes (similar to MUNIT). In SwapAE, the style encoder is forced to extract the global texture of the image by explicitly matching the patch statistics of the original image and swap-generated image (patch co-occurrence loss).

Training detail: batch 10, iteration 500k, single co-occurrence patch per sample

Samples

./imgs/swapae/1.png

Style code interpolation

./imgs/swapae/2.png

./imgs/swapae/3.png

./imgs/swapae/4.png

Structure code interpolation

./imgs/swapae/5.png

./imgs/swapae/6.png

./imgs/swapae/7.png

Simply interpolating the structure code didn't work well. The regional structure code editing method introduced in the original paper might work.

Swapping codes at different layers

./imgs/swapae/8.png

Injecting target style codes from the very first layer of the decoder often changes the whole identity of the original character. Detailed structures can be better preserved by applying the source style in the first few layers.

swap location = 2

./imgs/swapae/9.png

./imgs/swapae/10.png

./imgs/swapae/11.png

swap location = 3

./imgs/swapae/12.png

./imgs/swapae/13.png

./imgs/swapae/14.png

For the webcomics data domain, injecting target style codes from the 2nd~3rd layer gives pleasing style-transfer results.

swap location = 7

./imgs/swapae/15.png

./imgs/swapae/16.png

./imgs/swapae/17.png

The last few layers of the generator control the overall color.

Test image

From Slam Dunk

./imgs/swapae/18.png

Face2Webtoon

Finetuning SwapAE

Finetuning the webtoon-trained SwapAE model with real human faces for small amount of steps improves the robustness. Using the StyleGAN as a style image generator, exploration of styles without real webtoon data is also possible. See train_aux.py and face2webtoon.ipynb. An unconditional lightweight model can be found here.

./imgs/face2webtoon/dh.jpg

Facial cartoonizer + Style Swapping

Without finetuning, EdgeExtraction [code] + FacialCartoonization [code] → SwapAE also produces nice results.

./imgs/swapae/19.png

./imgs/swapae/20.png

./imgs/swapae/21.png

Failures

./imgs/swapae/22.png

JTBC Samples

./imgs/swapae/jtbc.png

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].