Interpreting the Latent Space of GANs for Semantic Face Editing
Authors: Yujun Shen, Jinjin Gu, Xiaoou Tang, Bolei Zhou Description: Despite the recent advance of Generative Adversarial Networks (GANs) in high-fidelity image synthesis, there lacks enough understanding of how GANs are able to map a latent code sampled from a random distribution to a photo-realistic image. Previous work assumes the latent space learned by GANs follows a distributed representation but observes the vector arithmetic phenomenon. In this work, we propose a novel framework, called InterFaceGAN, for semantic face editing by interpreting the latent semantics learned by GANs. In this framework, we conduct a detailed study on how different semantics are encoded in the latent space of GANs for face synthesis. We find that the latent code of well-trained generative models actually learns a disentangled representation after linear transformations. We explore the disentanglement between various semantics and manage to decouple some entangled semantics with subspace projection, leading to more precise control of facial attributes. Besides manipulating gender, age, expression, and the presence of eyeglasses, we can even vary the face pose as well as fix the artifacts accidentally generated by GAN models. The proposed method is further applied to achieve real image manipulation when combined with GAN inversion methods or some encoder-involved models. Extensive results suggest that learning to synthesize faces spontaneously brings a disentangled and controllable facial attribute representation.
Видео Interpreting the Latent Space of GANs for Semantic Face Editing канала ComputerVisionFoundation Videos
Видео Interpreting the Latent Space of GANs for Semantic Face Editing канала ComputerVisionFoundation Videos
Показать
Комментарии отсутствуют
Информация о видео
17 июля 2020 г. 13:55:19
00:01:01
Другие видео канала
![Exploring Style GAN2 Latent Vector: Controlling Facial Properties](https://i.ytimg.com/vi/5XX4uy9Mk9I/default.jpg)
![Bolei Zhou - Inverting Latent Space of GANs for Real Image Editings](https://i.ytimg.com/vi/zyBQ9obuqfQ/default.jpg)
![A.I. Experiments: Visualizing High-Dimensional Space](https://i.ytimg.com/vi/wvsE8jm1GzE/default.jpg)
![Isolating Sources of Disentanglements in VAEs | NeurIPS 2018](https://i.ytimg.com/vi/OK4f3bGQ4Bw/default.jpg)
![OpenAI’s CLIP explained! | Examples, links to code and pretrained model](https://i.ytimg.com/vi/dh8Rxhf7cLU/default.jpg)
![TL-GAN interface demo run 1](https://i.ytimg.com/vi/O1by05eX424/default.jpg)
![Editing Faces using Artificial Intelligence](https://i.ytimg.com/vi/dCKbRCUyop8/default.jpg)
![](https://i.ytimg.com/vi/xaKtLKmdTtc/default.jpg)
![Variational Autoencoders](https://i.ytimg.com/vi/9zKuYvjFFS8/default.jpg)
![MODALS: Modality-agnostic Automated Data Augmentation in the Latent Space](https://i.ytimg.com/vi/z_jZcdHrJxs/default.jpg)
![Latent Space Human Face Synthesis | Two Minute Papers #191](https://i.ytimg.com/vi/aR6M0MQBo2w/default.jpg)
![A GPT-3 for Images? Dall-E is the most impressive AI ever created!](https://i.ytimg.com/vi/xR6MxVoWg2E/default.jpg)
![CVPR'20 iMLCV tutorial: Exploring and Exploiting Interpretable Semantics in GANs by Bolei Zhou](https://i.ytimg.com/vi/rfx3whKgFVo/default.jpg)
![GPT-3 - explained in layman terms.](https://i.ytimg.com/vi/SboKeK6FFHQ/default.jpg)
![Latent Space](https://i.ytimg.com/vi/HcvTrvCntBY/default.jpg)
![AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks](https://i.ytimg.com/vi/Epvh4EvznUA/default.jpg)
![CLIP: Connecting Text and Images](https://i.ytimg.com/vi/u0HG77RNhPE/default.jpg)
![Visualizing Convolutional Filters from a CNN](https://i.ytimg.com/vi/cNBBNAxC8l4/default.jpg)
![The GAN That Warped: Semantic Attribute Editing With Unpaired Data](https://i.ytimg.com/vi/6bHmUakmVSA/default.jpg)
![Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling](https://i.ytimg.com/vi/mfx7uAkUtCI/default.jpg)