Google RealFill, an AI-powered generative image completion model, was spotted in a trademark listing

According to the business, Google RealFill can generate graphics despite significant discrepancies between references and targets

Google RealFill could be the tech giant’s next attempt to transform artificial intelligence (AI)-powered image production into a user-friendly application. Recently, a research paper and a website called RealFill were discovered online, which does image completion and inpainting based on reference photos and generates the target image. The company appears to have sought for trademarks on its created logo for the AI model-based product. Notably, the new AI model employs computer vision and pattern recognition algorithms and was trained with random masking approaches.

A Github page and a pre-print document of the AI model were just discovered by Android Authority. The magazine also discovered trademark applications filed under the name of Google LLC in the US Patent and Trade Office (USPTO) and European Union Intellectual Property Office (EUIPO) databases. According to them, the IT behemoth has not only completed the research phase for the AI model, but also intends to launch it as a commercial product.

RealFill’s GitHub page describes it as a “novel generative approach for image completion that fills in missing regions of an image with the content that should have been there.” Essentially, the AI model can scan numerous photographs of a human in the same environment and then utilize that reference to create a pre-specified image. As a tool, it can be utilized when a user takes many photos of an object but fails to capture the right shot. AI can process the photos and create an image that does not exist.

Read also:-Interview with Ruplekha Sinha roy

RealFill is a generative AI model that uses computer vision to grasp the subject and environment of reference pictures and can process a variety of characteristics of them, including technical parameters such as dimensions, colors, and forms, as well as contextual comprehension of the different objects. Using this information, it can then generate a target image in a new plane, filling in details that may not have been present in the reference photos.

While it is impossible to foresee Google’s plans for RealFill, last year the firm released a feature in the Google Pixel 8 series called greatest Take, which could process numerous images of a group photo and allow users to select the greatest expression from each image to generate the final snapshot. Compared to RealFill, the capability looks to be quite simple, but based on its application, the tech giant may be preparing the AI model to make it a feature for Pixel phones.