search
HomeTechnology peripheralsAIThe authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

Getting a great photo is getting easier and easier.

When traveling during the holidays, taking photos is indispensable. However, most of the photos taken in scenic spots are more or less regretful. Either there is something extra in the background, or there is something missing.

Obtaining a "perfect" image is one of the goals that CV researchers have long strived for. Recently, researchers from Google Research and Cornell University collaborated to propose an “Authentic Image Completion” technology—RealFill, a generative model for image completion.

The advantage of the RealFill model is that it can be personalized with a small number of scene reference images. These reference images do not need to be aligned with the target image, and can even be modified in terms of viewing angle, lighting conditions, camera aperture or image style. Big difference. Once personalization is complete, RealFill can complement the target image with visually engaging content in a way that is true to the original scene.

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

  • Please click the following link to view the paper: https://arxiv.org/abs/2309.16668

  • Project page link: https://realfill.github.io/

The inpainting and outpainting models are technologies that can generate high-quality and reasonable image content in unknown areas of the image, but the content generated by these models is necessarily unrealistic. , because these models have insufficient contextual information in real scenes. In contrast, RealFill generates content that "should" be there, making the results of image completion more realistic.

The authors pointed out in the paper that they defined a new image completion problem - "Authentic Image Completion". Different from traditional generative image restoration (the content that replaces the missing area may be inconsistent with the original scene), the goal of real image completion is to make the completed content as faithful as possible to the original scene, using content that "should appear there". Complete the target image with content that “might be out there.”

The authors stated that RealFill is the first method to extend the expressive power of generative image repair models by adding more conditions (i.e. adding reference images) to the process.

RealFill significantly outperforms existing methods on a new image completion benchmark covering a diverse and challenging set of scenarios.

Method

RealFill’s goal is to maintain as much authenticity as possible , using a small number of reference images to complete the missing parts of a given target image. Specifically, you are given up to 5 reference images, and a target image that roughly captures the same scene (but may have a different layout or appearance).

For a given scene, the researchers first create a personalized generative model by fine-tuning a pre-trained inpainting diffusion model on reference and target images. This fine-tuning process is designed so that the fine-tuned model not only maintains good image priors, but also learns the scene content, lighting, and style in the input image. This fine-tuned model is then used to fill in missing regions in the target image through a standard diffusion sampling process. The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

It is worth noting that for practical application value, this model pays special attention to more challenging, In the unconstrained case, the target image and the reference image may have very different viewpoints, environmental conditions, camera apertures, image styles, and even moving objects.

Experimental results

According to the reference image on the left, RealFill can The target image is expanded (uncrop) or repaired (inpaint), and the generated result is not only visually appealing, but also consistent with the reference image, even if the reference image and the target image differ in aspects such as viewpoint, aperture, lighting, image style, and object motion. There are big differences.

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFillThe authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFillThe output effect of RealFill model. Given a reference image on the left, RealFill can expand the corresponding target image on the right. The areas inside the white box are provided to the network as known pixels, while the areas outside the white box are generated. The results show that RealFill can generate high-quality images that are faithful to the reference image even if there are huge differences between the reference image and the target image, including viewpoint, aperture, lighting, image style, and object motion. Source: Paper

Controlled Experiment

The researchers compared the RealFill model with other baseline methods. In comparison, RealFill produces high-quality results and performs better in terms of scene fidelity and consistency with reference images.

Paint-by-Example cannot achieve a high degree of scene fidelity because it relies on CLIP embedding, which can only capture high-level semantic information.

Stable Diffusion Inpainting can produce seemingly reasonable results, but due to the limited expressive power of prompt, the final generated results are not consistent with the reference image.

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

Comparison of RealFill with two other baseline methods. The area covered by a transparent white mask is the unmodified portion of the target image. Source: realfill.github.io

##Limitations

Researchers also Some potential issues and limitations of the RealFill model are discussed, including processing speed, the ability to handle viewpoint changes, and the ability to handle situations that are challenging for the underlying model. Specifically:

RealFill requires a gradient-based fine-tuning process on the input image, which makes it relatively slow to run.
When the viewpoint change between the reference image and the target image is very large, RealFill is often unable to restore the 3D scene, especially when there is only one reference image.

Since RealFill mainly relies on image priors inherited from the base pre-trained model, it cannot handle situations that are challenging for the base model, such as stable The diffusion model does not handle text well.

The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill

Finally, the author expresses his gratitude to the collaborators:

We would like to thank Rundi Wu, Qianqian Wang, Viraj Shah, Ethan Weber, Zhengqi Li , Kyle Genova, Boyang Deng, Maya Goldenberg, Noah Snavely, Ben Poole, Ben Mildenhall, Alex Rav-Acha, Pratul Srinivasan, Dor Verbin, and Jon Barron for valuable discussions and feedback, and also thanks to Zeya Peng, Rundi Wu, Shan Nan for Evaluate the contribution of the dataset. We are especially grateful to Jason Baldridge, Kihyuk Sohn, Kathy Meier-Hellstern, and Nicole Brichtova for their feedback and support on the project.

Please read the original paper and visit the project homepage for more information

The above is the detailed content of The authenticity is shocking! Google and Cornell University launch real-life image completion technology RealFill. For more information, please follow other related articles on the PHP Chinese website!

Statement
This article is reproduced at:机器之心. If there is any infringement, please contact admin@php.cn delete
Gemma Scope: Google's Microscope for Peering into AI's Thought ProcessGemma Scope: Google's Microscope for Peering into AI's Thought ProcessApr 17, 2025 am 11:55 AM

Exploring the Inner Workings of Language Models with Gemma Scope Understanding the complexities of AI language models is a significant challenge. Google's release of Gemma Scope, a comprehensive toolkit, offers researchers a powerful way to delve in

Who Is a Business Intelligence Analyst and How To Become One?Who Is a Business Intelligence Analyst and How To Become One?Apr 17, 2025 am 11:44 AM

Unlocking Business Success: A Guide to Becoming a Business Intelligence Analyst Imagine transforming raw data into actionable insights that drive organizational growth. This is the power of a Business Intelligence (BI) Analyst – a crucial role in gu

How to Add a Column in SQL? - Analytics VidhyaHow to Add a Column in SQL? - Analytics VidhyaApr 17, 2025 am 11:43 AM

SQL's ALTER TABLE Statement: Dynamically Adding Columns to Your Database In data management, SQL's adaptability is crucial. Need to adjust your database structure on the fly? The ALTER TABLE statement is your solution. This guide details adding colu

Business Analyst vs. Data AnalystBusiness Analyst vs. Data AnalystApr 17, 2025 am 11:38 AM

Introduction Imagine a bustling office where two professionals collaborate on a critical project. The business analyst focuses on the company's objectives, identifying areas for improvement, and ensuring strategic alignment with market trends. Simu

What are COUNT and COUNTA in Excel? - Analytics VidhyaWhat are COUNT and COUNTA in Excel? - Analytics VidhyaApr 17, 2025 am 11:34 AM

Excel data counting and analysis: detailed explanation of COUNT and COUNTA functions Accurate data counting and analysis are critical in Excel, especially when working with large data sets. Excel provides a variety of functions to achieve this, with the COUNT and COUNTA functions being key tools for counting the number of cells under different conditions. Although both functions are used to count cells, their design targets are targeted at different data types. Let's dig into the specific details of COUNT and COUNTA functions, highlight their unique features and differences, and learn how to apply them in data analysis. Overview of key points Understand COUNT and COU

Chrome is Here With AI: Experiencing Something New Everyday!!Chrome is Here With AI: Experiencing Something New Everyday!!Apr 17, 2025 am 11:29 AM

Google Chrome's AI Revolution: A Personalized and Efficient Browsing Experience Artificial Intelligence (AI) is rapidly transforming our daily lives, and Google Chrome is leading the charge in the web browsing arena. This article explores the exciti

AI's Human Side: Wellbeing And The Quadruple Bottom LineAI's Human Side: Wellbeing And The Quadruple Bottom LineApr 17, 2025 am 11:28 AM

Reimagining Impact: The Quadruple Bottom Line For too long, the conversation has been dominated by a narrow view of AI’s impact, primarily focused on the bottom line of profit. However, a more holistic approach recognizes the interconnectedness of bu

5 Game-Changing Quantum Computing Use Cases You Should Know About5 Game-Changing Quantum Computing Use Cases You Should Know AboutApr 17, 2025 am 11:24 AM

Things are moving steadily towards that point. The investment pouring into quantum service providers and startups shows that industry understands its significance. And a growing number of real-world use cases are emerging to demonstrate its value out

See all articles

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

AI Hentai Generator

AI Hentai Generator

Generate AI Hentai for free.

Hot Article

R.E.P.O. Energy Crystals Explained and What They Do (Yellow Crystal)
1 months agoBy尊渡假赌尊渡假赌尊渡假赌
R.E.P.O. Best Graphic Settings
1 months agoBy尊渡假赌尊渡假赌尊渡假赌
R.E.P.O. How to Fix Audio if You Can't Hear Anyone
1 months agoBy尊渡假赌尊渡假赌尊渡假赌
R.E.P.O. Chat Commands and How to Use Them
1 months agoBy尊渡假赌尊渡假赌尊渡假赌

Hot Tools

mPDF

mPDF

mPDF is a PHP library that can generate PDF files from UTF-8 encoded HTML. The original author, Ian Back, wrote mPDF to output PDF files "on the fly" from his website and handle different languages. It is slower than original scripts like HTML2FPDF and produces larger files when using Unicode fonts, but supports CSS styles etc. and has a lot of enhancements. Supports almost all languages, including RTL (Arabic and Hebrew) and CJK (Chinese, Japanese and Korean). Supports nested block-level elements (such as P, DIV),

Safe Exam Browser

Safe Exam Browser

Safe Exam Browser is a secure browser environment for taking online exams securely. This software turns any computer into a secure workstation. It controls access to any utility and prevents students from using unauthorized resources.

ZendStudio 13.5.1 Mac

ZendStudio 13.5.1 Mac

Powerful PHP integrated development environment

SublimeText3 Mac version

SublimeText3 Mac version

God-level code editing software (SublimeText3)

Dreamweaver Mac version

Dreamweaver Mac version

Visual web development tools