Biotech & Health

Facebook’s new AI research is a real eye-opener

Comment

There are plenty of ways to manipulate photos to make you look better, remove red-eye or lens flare and so on. But so far the blink has proven a tenacious opponent of good snapshots. That may change with research from Facebook that replaces closed eyes with open ones in a remarkably convincing manner.

It’s far from the only example of intelligent “in-painting,” as the technique is called when a program fills in a space with what it thinks belongs there. Adobe in particular has made good use of it with its “context-aware fill,” allowing users to seamlessly replace undesired features, for example a protruding branch or a cloud, with a pretty good guess at what would be there if it weren’t.

But some features are beyond the tools’ capacity to replace, one of which is eyes. Their detailed and highly variable nature make it particularly difficult for a system to change or create them realistically.

Facebook, which probably has more pictures of people blinking than any other entity in history, decided to take a crack at this problem.

It does so with a Generative Adversarial Network, essentially a machine learning system that tries to fool itself into thinking its creations are real. In a GAN, one part of the system learns to recognize, say, faces, and another part of the system repeatedly creates images that, based on feedback from the recognition part, gradually grow in realism.

From left to right: “Exemplar” images, source images, Photoshop’s eye-opening algorithm and Facebook’s method.

In this case the network is trained to both recognize and replicate convincing open eyes. This could be done already, but as you can see in the examples at right, existing methods left something to be desired. They seem to paste in the eyes of the people without much consideration for consistency with the rest of the image.

Machines are naive that way: they have no intuitive understanding that opening one’s eyes does not also change the color of the skin around them. (For that matter, they have no intuitive understanding of eyes, color or anything at all.)

What Facebook’s researchers did was to include “exemplar” data showing the target person with their eyes open, from which the GAN learns not just what eyes should go on the person, but how the eyes of this particular person are shaped, colored and so on.

The results are quite realistic: there’s no color mismatch or obvious stitching because the recognition part of the network knows that that’s not how the person looks.

In testing, people mistook the fake eyes-opened photos for real ones, or said they couldn’t be sure which was which, more than half the time. And unless I knew a photo was definitely tampered with, I probably wouldn’t notice if I was scrolling past it in my newsfeed. Gandhi looks a little weird, though.

It still fails in some situations, creating weird artifacts if a person’s eye is partially covered by a lock of hair, or sometimes failing to recreate the color correctly. But those are fixable problems.

You can imagine the usefulness of an automatic eye-opening utility on Facebook that checks a person’s other photos and uses them as reference to replace a blink in the latest one. It would be a little creepy, but that’s pretty standard for Facebook, and at least it might save a group photo or two.

More TechCrunch

Maad, a B2B e-commerce startup based in Senegal, has secured $3.2 million debt-equity funding to bolster its growth in the western Africa country and to explore fresh opportunities in the…

Maad raises $3.2M seed amid B2B e-commerce sector turbulence in Africa

The fresh funds were raised from two investors who transferred the capital into a special purpose vehicle, a legal entity associated with the OpenAI Startup Fund.

OpenAI Startup Fund raises additional $5M

Accel has invested in more than 200 startups in the region to date, making it one of the more prolific VCs in this market.

Accel has a fresh $650M to back European early-stage startups

Kyle Vogt, the former founder and CEO of self-driving car company Cruise, has a new VC-backed robotics startup focused on household chores. Vogt announced Monday that the new startup, called…

Cruise founder Kyle Vogt is back with a robot startup

When Keith Rabois announced he was leaving Founders Fund to return to Khosla Ventures in January, it came as a shock to many in the venture capital ecosystem — and…

From Miles Grimshaw to Eva Ho, venture capitalists continue to play musical chairs

On the heels of OpenAI announcing the latest iteration of its GPT large language model, its biggest rival in generative AI in the U.S. announced an expansion of its own.…

Anthropic is expanding to Europe and raising more money

If you’re looking for a Starliner mission recap, you’ll have to wait a little longer, because the mission has officially been delayed.

TechCrunch Space: You rock(et) my world, moms

Apple devoted a full event to iPad last Tuesday, roughly a month out from WWDC. From the invite artwork to the polarizing ad spot, Apple was clear — the event…

Apple iPad Pro M4 vs. iPad Air M2: Reviewing which is right for most

Terri Burns, a former partner at GV, is venturing into a new chapter of her career by launching her own venture firm called Type Capital. 

GV’s youngest partner has launched her own firm

The decision to go monochrome was probably a smart one, considering the candy-colored alternatives that seem to want to dazzle and comfort you.

ChatGPT’s new face is a black hole

Apple and Google announced on Monday that iPhone and Android users will start seeing alerts when it’s possible that an unknown Bluetooth device is being used to track them. The…

Apple and Google agree on standard to alert people when unknown Bluetooth devices may be tracking them

The company is describing the event as “a chance to demo some ChatGPT and GPT-4 updates.”

OpenAI’s ChatGPT announcement: Watch here

A human safety operator will be behind the wheel during this phase of testing, according to the company.

GM’s Cruise ramps up robotaxi testing in Phoenix

OpenAI announced a new flagship generative AI model on Monday that they call GPT-4o — the “o” stands for “omni,” referring to the model’s ability to handle text, speech, and…

OpenAI debuts GPT-4o ‘omni’ model now powering ChatGPT

Featured Article

The women in AI making a difference

As a part of a multi-part series, TechCrunch is highlighting women innovators — from academics to policymakers —in the field of AI.

12 hours ago
The women in AI making a difference

The expansion of Polar Semiconductor’s facility would enable the company to double its U.S. production capacity of sensor and power chips within two years.

White House proposes up to $120M to help fund Polar Semiconductor’s chip facility expansion

In 2021, Google kicked off work on Project Starline, a corporate-focused teleconferencing platform that uses 3D imaging, cameras and a custom-designed screen to let people converse with someone as if…

Google’s 3D video conferencing platform, Project Starline, is coming in 2025 with help from HP

Over the weekend, Instagram announced that it is expanding its creator marketplace to 10 new countries — this marketplace connects brands with creators to foster collaboration. The new regions include…

Instagram expands its creator marketplace to 10 new countries

You can expect plenty of AI, but probably not a lot of hardware.

Google I/O 2024: What to expect

The keynote kicks off at 10 a.m. PT on Tuesday and will offer glimpses into the latest versions of Android, Wear OS and Android TV.

Google I/O 2024: How to watch

Four-year-old Mexican BNPL startup Aplazo facilitates fractionated payments to offline and online merchants even when the buyer doesn’t have a credit card.

Aplazo is using buy now, pay later as a stepping stone to financial ubiquity in Mexico

We received countless submissions to speak at this year’s Disrupt 2024. After carefully sifting through all the applications, we’ve narrowed it down to 19 session finalists. Now we need your…

Vote for your Disrupt 2024 Audience Choice favs

Co-founder and CEO Bowie Cheung, who previously worked at Uber Eats, said the company now has 200 customers.

Healthy growth helps B2B food e-commerce startup Pepper nab $30 million led by ICONIQ Growth

Booking.com has been designated a gatekeeper under the EU’s DMA, meaning the firm will be regulated under the bloc’s market fairness framework.

Booking.com latest to fall under EU market power rules

Featured Article

‘Got that boomer!’: How cybercriminals steal one-time passcodes for SIM swap attacks and raiding bank accounts

Estate is an invite-only website that has helped hundreds of attackers make thousands of phone calls aimed at stealing account passcodes, according to its leaked database.

16 hours ago
‘Got that boomer!’: How cybercriminals steal one-time passcodes for SIM swap attacks and raiding bank accounts

Squarespace is being taken private in an all-cash deal that values the company on an equity basis at $6.6 billion.

Permira is taking Squarespace private in a $6.9 billion deal

AI-powered tools like OpenAI’s Whisper have enabled many apps to make transcription an integral part of their feature set for personal note-taking, and the space has quickly flourished as a…

Buy Me a Coffee’s founder has built an AI-powered voice note app

Airtel, India’s second-largest telco, is partnering with Google Cloud to develop and deliver cloud and GenAI solutions to Indian businesses.

Google partners with Airtel to offer cloud and GenAI products to Indian businesses

To give AI-focused women academics and others their well-deserved — and overdue — time in the spotlight, TechCrunch has been publishing a series of interviews focused on remarkable women who’ve contributed to…

Women in AI: Rep. Dar’shun Kendrick wants to pass more AI legislation

We took the pulse of emerging fund managers about what it’s been like for them during these post-ZERP, venture-capital-winter years.

A reckoning is coming for emerging venture funds, and that, VCs say, is a good thing