Adobe, Inc.

Other

Princeton University

Czech Technical University in Prague

Jingwan Lu

Elya Shechtman

Richard Zhang

Yijun Li

Shabnam Ghadar

Ratheesh Kalarot

Zhe Lin

Jianming Zhang

Paul John Asente

Wei-An Lin

Krishna Kumar Singh

Cameron Smith

Scott David Cohen

Baldo Antonio Faieta

Haitian Zheng

Ning Xu

Jakub Fiser

Stephen J DiVerdi

Sohrab Amirghodsi

Zhixin Shu

Jun-Yan Zhu

Ondrej Jamriska

Connelly Barnes

Niloy Jyoti Mitra

Qing Liu

Duygu Ceylan Aksit

Daniel Sýkora

Alexei A Efros

Michal Lukác

Oliver Wang

Jimei Yang

Jose Ignacio Echevarria Vallespi

David P Simons

Aliakbar Darabi

Tobias Hinz

Radomir Mech

Matthew David Fisher

Adam Finkelstein

Akhilesh Kumar

Yangtuanfeng Wang

John Thomas Nack

Elya Schechtman

Yuheng Li

Sachin Soni

Nathan Aaron Carr

Ajay Jain

Byungmoon Kim

Chen Fang

Gautham J Mysore

Sanjeev Tagra

Federico Perazzi

Holger Winnemoeller

Kevin Wampler

Zhifei Zhang

Prasenjit Mondal

Michal Lukac

Christian Thomas Cantrell

Taesung Park

Zeyu Jin

Junyan Zhu

Hui Qu

Alexandru Vasile Costin

Wilson Chan

Maria Shugrina

Archie Samuel Bagnall

Patrick James Hebron

Sven Olsen

Gaurav Parmar

Piotr Walczyszyn

Timothy M Converse

Yannick Hold-Geoffrey

Jun Xing

Alannah Oleson

Jose Echevarria

Patsorn Sangkloy

Ajay Bedi

Vineet Batra

Ankit Phogat

Li-Yi Wei

Joon-Young Lee

Yang Yang

Saeid Motiian

Yannick Hold-Geoffroy

Yi Zhou

Sameer Bhatt

Yuqian Zhou

Mridul Kavidayal

Alex Charles Filipkowski

Stephen Diverdi

Qi Sun

Nishant Kumar

Vikas Sharma

Hung-Yu Tseng

Julia Gong

Nicholas Isaac Kolkin

Seoungwug Oh

Rebecca Ilene Milman

Huiwen Chang

Systems and methods for image processing are described. Embodiments of the present disclosure obtain a noise image and guidance information for generating an image. A diffusion model generates an intermediate noise prediction for the image based on the noise image. A conditioning network generates noise modulation parameters. The intermediate noise prediction and the noise modulation parameters are combined to obtain a modified intermediate noise prediction. The diffusion model generates the image based on the modified intermediate noise prediction, wherein the image depicts a scene based on the guidance information.

Multimodal diffusion models

The present disclosure relates to systems, methods, and non-transitory computer-readable media that modify digital images via scene-based editing using image understanding facilitated by artificial intelligence. For example, in one or more embodiments the disclosed systems utilize generative machine learning models to create modified digital images portraying human subjects. In particular, the disclosed systems generate modified digital images by performing infill modifications to complete a digital image or human inpainting for portions of a digital image that portrays a human. Moreover, in some embodiments, the disclosed systems perform reposing of subjects portrayed within a digital image to generate modified digital images. In addition, the disclosed systems in some embodiments perform facial expression transfer and facial expression animations to generate modified digital images or animations.

Synthesizing a modified digital image utilizing a reposing model

Systems and methods for image processing are described. Embodiments of the present disclosure receive an image comprising a first region that includes content and a second region to be inpainted. Noise is then added to the image to obtain a noisy image, and a plurality of intermediate output images are generated based on the noisy image using a diffusion model trained using a perceptual loss. The intermediate output images predict a final output image based on a corresponding intermediate noise level of the diffusion model. The diffusion model then generates the final output image based on the intermediate output image. The final output image includes inpainted content in the second region that is consistent with the content in the first region.

Image and object inpainting with diffusion models

The present disclosure relates to systems, methods, and non-transitory computer readable media that generate inpainted digital images utilizing class-specific cascaded modulation inpainting neural network. For example, the disclosed systems utilize a class-specific cascaded modulation inpainting neural network that includes cascaded modulation decoder layers to generate replacement pixels portraying a particular target object class. To illustrate, in response to user selection of a replacement region and target object class, the disclosed systems utilize a class-specific cascaded modulation inpainting neural network corresponding to the target object class to generate an inpainted digital image that portrays an instance of the target object class within the replacement region. Moreover, in one or more embodiments the disclosed systems train class-specific cascaded modulation inpainting neural networks corresponding to a variety of target object classes, such as a sky object class, a water object class, a ground object class, or a human object class.

Object class inpainting in digital images utilizing class-specific inpainting neural networks

Systems and methods train an encoder neural network for fast and accurate projection into the latent space of a Generative Adversarial Network (GAN). The encoder is trained by providing an input training image to the encoder and producing, by the encoder, a latent space representation of the input training image. The latent space representation is provided as input to the GAN to generate a generated training image. A latent code is sampled from a latent space associated with the GAN and the sampled latent code is provided as input to the GAN. The GAN generates a synthetic training image based on the sampled latent code. The sampled latent code is provided as input to the encoder to produce a synthetic training code. The encoder is updated by minimizing a loss between the generated training image and the input training image, and the synthetic training code and the sampled latent code.

Supervised learning techniques for encoder training

The present disclosure relates to systems, methods, and non-transitory computer readable media for panoptically guiding digital image inpainting utilizing a panoptic inpainting neural network. In some embodiments, the disclosed systems utilize a panoptic inpainting neural network to generate an inpainted digital image according to panoptic segmentation map that defines pixel regions corresponding to different panoptic labels. In some cases, the disclosed systems train a neural network utilizing a semantic discriminator that facilitates generation of digital images that are realistic while also conforming to a semantic segmentation. The disclosed systems generate and provide a panoptic inpainting interface to facilitate user interaction for inpainting digital images. In certain embodiments, the disclosed systems iteratively update an inpainted digital image based on changes to a panoptic segmentation map.

Generating and providing a panoptic inpainting interface for generating and modifying inpainted digital images

Iteratively modifying inpainted digital images based on changes to panoptic segmentation maps

Learning parameters for neural networks using a semantic discriminator and an object-level discriminator

This disclosure describes one or more implementations of a digital image semantic layout manipulation system that generates refined digital images resembling the style of one or more input images while following the structure of an edited semantic layout. For example, in various implementations, the digital image semantic layout manipulation system builds and utilizes a sparse attention warped image neural network to generate high-resolution warped images and a digital image layout neural network to enhance and refine the high-resolution warped digital image into a realistic and accurate refined digital image.

Generating digital images utilizing high-resolution sparse attention and semantic layout manipulation neural networks

Human inpainting utilizing a segmentation branch for generating an infill segmentation map

An improved system architecture uses a Generative Adversarial Network (GAN) including a specialized generator neural network to generate multiple resolution output images. The system produces a latent space representation of an input image. The system generates a first output image at a first resolution by providing the latent space representation of the input image as input to a generator neural network comprising an input layer, an output layer, and a plurality of intermediate layers and taking the first output image from an intermediate layer, of the plurality of intermediate layers of the generator neural network. The system generates a second output image at a second resolution different from the first resolution by providing the latent space representation of the input image as input to the generator neural network and taking the second output image from the output layer of the generator neural network.

Multi-scale output techniques for generative adversarial networks

The present disclosure relates to systems, non-transitory computer-readable media, and methods for utilizing machine learning models to generate modified digital images. In particular, in some embodiments, the disclosed systems generate image editing directions between textual identifiers of two visual features utilizing a language prediction machine learning model and a text encoder. In some embodiments, the disclosed systems generated an inversion of a digital image utilizing a regularized inversion model to guide forward diffusion of the digital image. In some embodiments, the disclosed systems utilize cross-attention guidance to preserve structural details of a source digital image when generating a modified digital image with a diffusion neural network.

Utilizing cross-attention guidance to preserve content in diffusion-based image modifications

Systems and methods train and apply a specialized encoder neural network for fast and accurate projection into the latent space of a Generative Adversarial Network (GAN). The specialized encoder neural network includes an input layer, a feature extraction layer, and a bottleneck layer positioned after the feature extraction layer. The projection process includes providing an input image to the encoder and producing, by the encoder, a latent space representation of the input image. Producing the latent space representation includes extracting a feature vector from the feature extraction layer, providing the feature vector lo the bottleneck layer as input, and producing the latent space representation as output. The latent space representation produced by the encoder is provided as input to the GAN, which generates an output image based upon the latent space representation.

Direct regression encoder architecture and training

An image inpainting system is described that receives an input image that includes a masked region. From the input image, the image inpainting system generates a synthesized image that depicts an object in the masked region by selecting a first code that represents a known factor characterizing a visual appearance of the object and a second code that represents an unknown factor characterizing the visual appearance of the object apart from the known factor in latent space. The input image, the first code, and the second code are provided as input to a generative adversarial network that is trained to generate the synthesized image using contrastive losses. Different synthesized images are generated from the same input image using different combinations of first and second codes, and the synthesized images are output for display.

Diverse image inpainting using contrastive learning

Generating a modified digital image utilizing a human inpainting model

Growing community of inventors

Santa Clara, CA, United States of America

Jingwan Lu