跳到正文
原文
Google 博客 · XR/AR·· 2025-01-28AI 评分40
光学与 AI

Google讲述Pixel相机Add Me合影功能的开发过程

How we built Pixel’s Add Me feature for easier group photos

主题与目录
AI 导读

Google介绍Pixel相机Add Me功能的研发经过,该功能将同一场景中先后拍摄的两张照片合成,让拍照者也能出现在合影中。第二位拍摄者会看到第一张照片的增强现实叠加层以对齐构图,功能依靠多个机器学习模型,在Tensor G4芯片的TPU上本地运行。Add Me随Pixel 9系列推出。

正文 · 原文

中文译文尚未生成,当前展示的是已获取的原文文本。

Jan 27, 2025

|

Say cheese, and Add Me please!



Maayan Rossmann is her family’s de facto photographer, which means she’s missing from a lot of photos. “In Israel, there’s ‘Family Day,’ where kids bring in a family photo at school,” the Pixel camera product manager says. “And it was always hard for me to find one that I was in!”

This experience, of course, isn’t specific to parents; there’s always someone who volunteers to take the pic, or dash for the camera timer or contort their arm to get everyone in the shot. But the plight of the parent behind the lens is a big part of what inspired Pixel’s Add Me feature. Launched at Made by Google in August, Add Me allows you to combine two pictures taken during the same session and in the same scene so everyone — including the person who took the original shot — is in the photo.

With Add Me, you’ll get a photo with everyone who was there — without having to pack a tripod or ask a stranger for help.

Software engineer Adi Zicher had only been on the Creative Camera team for about three weeks in August 2022 before pitching the idea. During a brainstorm session, Adi brought up that someone is always left out of group shots. This resonated with everyone, especially the parents (who, given their day jobs, are often the group photographer).

Development was already underway for the Pixel 8 series, so a feature that addressed this problem wouldn’t be ready in time. But Adi could still chip away at it for the future. “I had a year of working on it solo, and popping into hackathons and stuff, trying to get people to work on it with me,” Adi explains. Towards the end of that year, a mockup of the feature was available, which was also around the time Maayan joined the project. “Maayan took care of making this a reality!” Adi says.

Maayan and Adi using the Add Me feature.

Making Add Me a reality required collaboration between the Pixel Camera, Creative Camera and Google XR teams. (The Google XR team works on Android XR and ARCore, platforms for building augmented and virtual reality experiences.) To blend an image, Add Me uses augmented reality (AR) to show the second photographer an overlay of the first image so they can accurately frame the new photo to match the composition of the first one. Since aligning the two images is crucial to get the best result, the team explored a few methods to get the perfect shot. They eventually realized augmented reality was the most useful way to visually guide the user, so they turned to the Google XR team.

“We on the XR team have always wanted to do more and enable helpful features in the Camera app,” says XR software engineer Ryan DuToit. The team saw the vision for Add Me and recognized how useful it would be for personal photography.

Still, developing an interface where the AR feature was self-explanatory — so even those unfamiliar with the technology could use it — wasn’t easy, and took ample experimentation. Plus, the XR team had a busy 2024, Ryan explains, pointing to the recent Android XR launch. “Collaborating with another team on another project was challenging — but we did it, and I’m very proud of it,” he says.

Add Me was a fairly complicated technical feat: It uses various machine learning models, which are powered by Pixel’s TPU and the Tensor G4 chip to run on device. “If it weren’t for the TPU, I don’t think we would have been able to converge to a reasonable latency,” Adi says. If they’d had to run these ML models on a GPU or CPU, the feature wouldn’t have been able to display the images from the first shot and then later the blended shot quickly enough.

Related Article

What's the difference between a CPU, GPU and TPU?

Learn more from a Google expert about CPUs, GPUs and TPUs — and Google latest TPU, Trillium.

See more

Three squares illustrate different computer processors. Blue: a classic CPU with a grid of contact points. Green: a GPU with a simple line symbolizing parallel processing. Yellow: a TPU with intricate circuitry for machine learning.

Add Me launched with our Pixel 9 phones and quickly became a popular feature, especially during moments when people tend to gather. "When I look at the photos Googlers share with us, it strikes me how often they use Add Me to capture those special moments with loved ones — family gatherings, vacations, holidays,” Adi says.

And then there’s the whole “twinning with yourself” thing. With Add Me, you can have someone take two photos of you to blend together into interesting, Sci-Fi-esque “twinning” photographs. It’s something the team experimented with a lot — but not for the cool effect.

Adi twinning, thanks to Add Me.

A woman in a black jacket stands on the side of a cliff next to a suspension bridge. Another image of the woman is on the bridge.

“We’d take photos duplicating ourselves just because that was easy for testing,” Adi says. But the twinning craze took over the Sandbox demo at Made by Google. “It was one of the first things that people tried with it!” she says.

So whether you want to get in the next family photo or take one with the person you know best — yourself — Add Me can be the perfect tool.

来源:Google 博客 · XR/AR · blog.google