Click the card below to follow the " CVer " public account
AI/CV heavy dry goods, delivered as soon as possible
Xiao Xiao from Concave Temple
Reprinted from: Qubit (QbitAI)
Still going to the barbershop to get a job and shave your bald head? Try this generator!
As long as a photo is input, the output is a flawless bald head, and the temperament is immediately promoted from intern to supervisor (manual dog head)
Whether you are a young programmer or a programmer with fluttering hair, you can "become stronger" by this method:
With this "bald head artifact" named HairMapper , P hair is a matter of minutes, but the facial features are completely unaffected:
Well, if I have a hairstyle I want to do in the future, I will directly post one...
To achieve such a good effect, it is necessary to use three GANs to work together.
The "bald artifact" that everyone can use
Specifically, the principle of HairMapper is divided into three steps, generating baldness → retaining facial features → merging avatars.
First, the authors used StyleGAN to create a bald effect similar to the original head shape:
At this time, we saw that StyleGAN could not accurately control the facial features, so the face shape changed slightly.
Then, the next step is to use InterFaceGAN to cut out a face and facial features other than hair, while preserving the surrounding scenery:
Finally, combine the effects generated in the first two steps to make a perfect "bald head":
However, it is worth mentioning that due to reasons such as long hair and the lack of female bald head data sets, the authors added an extra step for the female version of the "bald head artifact".
In this, female photos must first be converted into male photos through StyleFlow, and then repeat the above steps, and finally get a bald head:
This effect is still good. Tests in the data set show that even girls with amazing hair volume can lose hair with one click (doge).
The authors compared HairMapper with other bald generators.
Also made a 6000+ bald data set
In fact, it's not that no one has done a bald generator before, but the effect is not ideal.
For example, Adobe's previous generator called StyleFlow can make Musk bald:
But one is that the hair is not clean, and the other is that the facial features have become "another person" for GAN. In addition to StyleFlow, other types of "bald generators" also face this problem:
On the basis of GAN hair loss, this HairMapper retains the original facial features, making the restored face very real.
Regardless of gender, age, or round and pointed tip:
Or a variety of skin tones and even different skin types:
Even the side face and even shadows can get silky bald special effects under the action of HairMapper:
The authors have used this algorithm to turn 6,000 portraits in the Flickr-Faces-HQ (FFHQ) dataset into bald heads.
In the future, whether it is P hair or a big man in men's/women's clothing, you can use this open source data set to train:
about the author
The first author, Yiqian Wu, is from the Computer Aided Design and Graphics (CAD&CG) Laboratory of Zhejiang University. He is currently a doctoral student with interests in computer vision and portrait editing.
The second author, Yong-Liang Yang, is an assistant professor at the University of Bath. He graduated from Tsinghua University with a bachelor’s, master’s and doctorate degree and studied under Professor Hu Shimin. His research interests are computer graphics, geometric modeling, VR/AR, etc.
The corresponding author, Jin Xiaogang, is a professor at the CAD&CG Laboratory of Zhejiang University. His research interests include movie special effects simulation, autonomous driving simulation, and creative modeling.
Can't wait to try the "bald artifact"?
According to the author, the pre-training code of HairMapper is still in urgent production and will be released soon, and the data set is already available~
Project address:
https://github.com/oneThousand1000/non-hair-FFHQ
Paper address:
http://www.cad.zju.edu.cn/home/jin/cvpr2022/HairMapper.pdf
ICCV and CVPR 2021 Paper and Code Download
Backstage reply: CVPR2021, you can download the CVPR 2021 papers and open source papers collection
Background reply: ICCV2021, you can download the ICCV 2021 papers and open source papers collection
Background reply: Transformer review, you can download the latest 3 Transformer reviews PDF
目标检测和Transformer交流群成立
扫描下方二维码，或者添加微信：CVer6666，即可添加CVer小助手微信，便可申请加入CVer-Transformer或者目标检测 微信交流群。另外其他垂直方向已涵盖：目标检测、图像分割、目标跟踪、人脸检测&识别、OCR、姿态估计、超分辨率、SLAM、医疗影像、Re-ID、GAN、NAS、深度估计、自动驾驶、强化学习、车道线检测、模型剪枝&压缩、去噪、去雾、去雨、风格迁移、遥感图像、行为识别、视频理解、图像融合、图像检索、论文投稿&交流、PyTorch、TensorFlow和Transformer等。
一定要备注：研究方向+地点+学校/公司+昵称（如Transformer或者目标检测+上海+上交+卡卡），根据格式备注，可更快被通过且邀请进群
▲扫码或加微信: CVer6666，进交流群
CVer学术交流群（知识星球）来了！想要了解最新最快最好的CV/DL/ML论文速递、优质开源项目、学习教程和实战训练等资料，欢迎扫描下方二维码，加入CVer学术交流群，已汇集数千人！
▲扫码进群
▲点击上方卡片，关注CVer公众号
整理不易，请点赞和在看