FaceApp's AI Facial Transformation Feature

From SI410
Jump to: navigation, search

FaceApp's Facial Transformation Features are features within the app, FaceApp. These features use artificial intelligence to create a modified version of a face in a photo. This feature can modify age, gender, and can do high quality 'airbrushing' of a photo. The app has been around since 2017, created by Russian developer Yaroslav Goncharov. These features of the app have been received both in positive and negative ways among users in areas of personal information and social media. FaceApp has had over 80 million users since 2018. [1]

Source: Google Play Store
Back • ↑Topics • ↑Categories

Features

FaceApp uses artificial intelligence to airbrush a photo. The artificial intelligence in the app, FaceApp, detects faces. After detecting a face, artificial intelligence is used to target parts of the face chosen by a user to create a more clear, airbrushed photo. FaceApp's artificial intelligence features are involved in several functions within FaceApp.

The category, retouch, in FaceApp prompts a user to draw over an area they would like to airbrush. Then, the artificial intelligence in FaceApp detects what part of the face is being drawn over, processes the image, and returns an airbrushed version to the user.

In the gender category, the app gives four choices in total, named 'masculine' or 'feminine' followed by numbers 1 or 2. As the number increases next to the gender selection, the amount of airbrushing done on the photo increases. This feature of FaceApp makes a face in an uploaded photo to appear either more feminine or masculine
Faceappfeature.png
The feminine feature produces a photo where the user's hair is longer and of brighter color. The skin on the face in the uploaded photo will appear more smooth and if teeth are in the photo, they will appear more white. The eyes on the face will be brightened and eyelashes will be lengthened, darkened. Makeup will be added to the photo. The face will have a smaller nose and more prominent cheekbones. The person in the photo will have more prominent collar bones and their upper half will be accentuated. In the masculine feature, the user's hair will be shorter, brighter, and facial hair will be added. The skin on the face will appear smooth and teeth will appear more white. The eyes on the face will be brightened but makeup will be removed. The face will have a wider nose and prominent jawline.

Within the gender feature, a user can change their gender entirely in a photo by artificial intelligence if the app is prompted to do so. FaceApp's artificial intelligence will detect a user's features that associate with the gender in the photo and change detected features to those that associate with the opposite gender.

In the age feature of FaceApp, a user can age their selfies. FaceApp detects the face and adds wrinkles in the forehead and mouth area. The feature also adds drooping skin around the eyes, jawline, and cheeks. It changes the user’s hair to be gray in color and skin to be of slightly paler color.

History

FaceApp was created in 2014 and released in 2017 for IOS and Android by Yaroslav Goncharov’s company Wireless Lab. The company is based in Russia in a tech hub west of Moscow, but is also incorporated inside the US in Delaware [2]. FaceApp’s founder worked at Windows Mobile for Microsoft, SPB Software and co-founded the Russian company named Yandex. As a part of Yandex, Goncharov headed their mobile platform department from 2011 to 2013. Prior to Yandex, Goncharov worked for SPB Software making home-screen replacements for Android devices. FaceApp’s founder worked for Microsoft before this[3]. By selling Yandex, Goncharov became very wealthy. Goncharov began FaceApp when he was forty-years-old with a staff of twelve. Many of FaceApp’s staff have similar professional backgrounds to Goncharov, working for Yandex, a Russian social-media platform named VKonatakte, and SPB Software.[4] FaceApp became popular amongst Americans after celebrities posted photos using the aging feature of the app [5]. FaceApp has attracted attention specifically from the LGBTQ community, specifically the transgender community because of the app’s ability to simulate a person in a photo as their opposite birth gender. FaceApp has attracted attention from instagram influencers due to its ability to airbrush a photo. FaceApp has more than 80 million users worldwide.[6]

Privacy

FaceApp’s servers are mostly based within the United States, hosting records in the United States, Ireland, and Singapore. After a user uploads a photo to FaceApp, the app uploads their photos to a server ‘cloud’ run by both Amazon and Google[7]. The app’s privacy policy states that the app can collect any photo uploaded by a user and data on websites visited by the user[8]. The app’s terms of service states the app has a license to use the user's photos, name, and likeness in any way the app sees fit. FaceApp removes photos from its servers after 24 to 48 hours in accordance with the app’s “perpetual” license to user’s photos. To improve the app, it must store data so that it can continue to train the algorithm that makes the app work. Connecting a user’s social media to the app allows for data-scraping by third parties[9]. FaceApp does not have access to every photo in a user’s camera roll. You can use FaceApp without giving your real name or email. For Russian users: users may have data stored in Russia. To delete data, users must make an inquiry in the “Report a Bug” tab within FaceApp. The app’s founder is currently working on a better user interface to delete user data[10].

Artificial Intelligence

To manipulate images, FaceApp uses image recognition technology. The image recognition technology allows facial recognition systems within the app to work. The technology uses deep learning to recognize facial features like eyes, nose, mouth, and jawlines. The machine learning model works on sample data that is gathered from a user’s device. The sample data that is collected is composed of the user’s images on their device that they choose to share with the app. By collecting the sample data, FaceApp’s artificial intelligence software consists of Generative Adversarial Networks (GANs), a deep-learning-based network, so that it can understand the human face in depth. The group of networks used for artificial intelligence in FaceApp are also used for high-fidelity natural image synthesis, augmenting data, and enhancing image compression. The Generative Adversarial Network associated with FaceApp uses two neural network algorithms that conflict with each other and eventually produce new information, a manipulated image. They are designed to recognize patterns on their own and manipulate images based on images pushed to the app. Its database of user images consists of millions of photos. In FaceApp, by using Generative Adversarial Networks for facial alteration, it often happens that the deep neural networks in the software lose the input information (the image) in the conflicting algorithm process which edits the images and produces entirely new information, a new image, for the user. To make the Generative Adversarial Networks more accurate, FaceApp uses a GAN that recognizes age and uses it as a conditional. The software works to preserve the age of the face in the pushed image so that an entirely new image is not created, unless prompted[11].

To create the opposite gender feature, similar artificial intelligence is used. Along with an age parameter, FaceApp uses gender-labelled data as a conditional and swaps the image to be that of the opposite. The sample data allows the app to manipulate a face with precision, the more photos a user adds of a face in FaceApp, the more accurately it will manipulate [12].

Social Media

Age Filter

In 2019 FaceApp released a new facial modification filter that made users look either younger or older than they, in reality, are. #FaceAppChallenge became popular across the internet, with many celebrities participating. To create a photo of a user at a different
Age.png
age, users select either young or old among the free features in FaceApp. An article by PopBuzz argues that the filter became viral because of its realistic qualities [13]. FaceApp’s artificial intelligence network has been heavily tested and refined by 2019 allowing for accurate facial transformations in the app. The age filter has been concerning for some members of the LGBTQ community for those who identify other than their assigned birth gender. An interviewee for an article by Allee Manning for Vocativ finds the age filter worrisome for their genderqueer/trans-feminine identity. They argue that the age filter makes them concerned to age because the results of the age filter on FaceApp make genderqueer users look more like their gender assigned at birth. They find this anxiety inducing as they age and that they must work harder to be seen as who they identify [14].

LGBTQ

"Yassify"

2021 brought a lot of new slang to the forefront of American culture, specifically ‘yassifying’ as dubbed by the LGBTQ communities. NBC News called the “yassification” of the internet being the meme that perhaps best embodies 2021[15]. By definition, yassifying is the act of beautifying something that is, typically, not appealing or heteronormative. The word “yassify” has a positive connotation among the LGBTQ communities. Being “yassified” has been used both as a slang word, describing something attractive and glamorous; or to describe extreme editing, typically done in FaceApp.
Yassify.png
Journalist Shane O’Neill from the New York Times explains examples of yassifying, “‘Girl With a Pearl Earring’ in a full face of makeup. The first Queen Elizabeth contoured from her neck ruff up. Severus Snape with jet-black hair extensions. Sasquatch sporting a smokey eye.” [16] The way this is done is through FaceApp’s gender feature. To modify an image to be “yassified”, one must apply the gender filter over the same photo multiple times. This works because every time the photo is uploaded to FaceApp, the more modifications on the photo that are done by FaceApp’s artificial intelligence. When applying the gender filter multiple times, the machine learning system within FaceApp looks for parts of the face to airbrush, rather than change the gender. This causes an image to become almost irrecognizable, and heavily made up, as FaceApp’s artificial intelligence networks lose the data that preserves the original image, creating a new one.
  1. See: Features
It was popularized in 2021 to “yassify" public or classical figures. Bringing “yassify” mainstream can be credited to the Twitter account named Yassify Bot, created by a 22-year-old college student in Omaha, Nebraska[17]. This Twitter account has used many photos of both historic and contemporary popular figures that have been heavily processed through FaceApp’s facial modification features. These transition photos that display a before and after photos on Twitter have received hundreds of thousands of likes and retweets[18].
Yassifytweet.png
Yassification has been noted to have similarities with drag, creating images that sometimes look fake and entirely overdone- similar to bimbofication. Recently, questions have been raised about ‘yassification’. It has been argued that the line between celebration and mocking is becoming too diffused, some images appearing completely ridiculous to some users. Other critical participants of “yassification” describe the trend as shallow and destructive[19].



Gender-Swap Feature

FaceApp’s gender-swapping feature has been both criticized and praised among LGBTQ users. In the gender-swapping feature, FaceApp uses artificial intelligence to swap the gender of a face in a user’s uploaded photo.

  1. See: Features

Critics say that the gender-swap feature in FaceApp has brought up difficult and strong emotions among LGBTQ users [20]. LGBTQ users who identify not as how they were assigned at birth are worried about using FaceApp’s gender swap feature because they are concerned about FaceApp showing them a face that is unattainable, yet desired. Transgender users argue that the gender-swapping feature makes transitioning appear as silly and that this FaceApp feature makes light of their experience [21]. Other users have had a more positive experience with FaceApp’s gender swap feature, especially members of the younger transgender communities. An interviewee for an article by Allee Manning for Vocativ recalls that they have had therapeutic experiences with FaceApp’s gender swap feature. They say that since they are a minor and are unable to get access to hormones, FaceApp helps them to deal with their gender dysmorphia. Though they see themselves as their gender assigned at birth in appearance, FaceApp has allowed them to see how they will look once old enough to transition to their desired identity [22]. Other users who share similarities with Manning’s interviewee say that the FaceApp gender-swapping features are affirming for them and are inspiring them to take their next steps in their transition journey. Other LGBTQ users say that they are okay with the use of FaceApp gender swapping features as long as transgender and non-binary people are treated the same as their cis-gender friends who have been making light of the app’s feature [23].

Criticisms

Racism

FaceApp’s artificial intelligence software has been criticized for lightening user’s skin tones in their gender feature, dubbed “hot” filter by the New York Times [24]. In an interview conducted by The Guardian, CEO of FaceApp, Yaroslav Goncharov apologized for FaceApp’s skin lightening that was occurring in this feature. Goncharov explained that the skin lightening was completely unintentional and that it developed from a bug in the app. In more technical terms, the bug developed from an underlying neural network caused by the images that were uploaded to the app for machine learning purposes- that skin lightening was no intention of the app’s creators [25]. On the other end of the spectrum, FaceApp has been criticized for allowing users to modify pictures to look like other races, journalist Alex Hern from The Guardian describing the feature allowed by FaceApp to be ‘digital blackface’ [26]. Users of FaceApp argue that the development of this feature in FaceApp was not sensitively handled and allows users to treat race as a ‘game of dress-up’. FaceApp’s owners, Wireless Lab, released a statement arguing that the ethnicity filters were designed to be equal amongst all races and that there are not positive or negative connotations associated with the feature. FaceApp’s owner, Yaroslav Goncharov, argues that the feature is equal for each race in the way that they are all represented by the same icon within the app and that every race listed in the app has the same list of filters. This feature of FaceApp’s artificial intelligence face modification abilities has been removed since 2017.

Personal Information Collection

Users have been skeptical about how FaceApp uses photos after processing them through artificial intelligence to create a modified version of themselves because of the app’s Russian origin. Some critics believe FaceApp to be sending users’ faces to the Russian government. In December the FBI issued a warning about apps based in Russia, including FaceApp, stating apps developed in Russia are a “potential counterintelligence threat” [27]. The app has a legalese critics call ‘vague’, critics further arguing that the app may be a disinformation campaign or downloading a user’s entire photo album without consent. FaceApp has caught the attention of leaders of the Democratic party who are warning campaign staff to delete the app because of how Russian hackers targeted DNC officials during the 2016 presidential race. FaceApp’s privacy terms do not guarantee a users’ data or information to be secure. The company behind FaceApp is allowed to share user information with third parties like companies and advertisers. If given a search warrant, subpoena, or a court order, FaceApp will share app data with a government agency. This is allowed in any country where FaceApp has an office. By using the app, users consent to transfer, processing, and storage of information in all countries, even though they may not have the same rights and protections under law in countries outside of the United States. FaceApp has a history of sharing information (any photo uploaded by users) with Facebook and Google AdMob[28]. This allows FaceApp and advertising companies to make money based on creating targeted ads through user's personal information. In June 2020, Forbes Senior Cybersecurity Contributor, Kate O’Flaherty, reassessed FaceApp’s privacy policy. FaceApps’s policies have since been rewritten to explicitly include that photos uploaded to the app are only used for the app, within the app to improve its neural network and fix bugs [29].

Removing Data

FaceApp’s founder, Yaroslav Goncharov, says that a user can remove their data from FaceApp from within the app. He instructs that users can make this request by reporting a bug to the app. More specifically, click through the app by selecting “Settings”, “Support”, “Report a Bug”, and then report the bug with “privacy” in the subject line [30]. FaceApp users say the process to request to have personal information deleted is convoluted. FaceApp does not have enough employees to respond to user’s requests to have their personal data removed from the app. FaceApp’s founder says they are working on a better user interface to remove data from the app, at user’s request. [31]

References

  1. The AI Behind FaceApp. (2020). Analyticsindiamag.Com. https://analyticsindiamag.com/the-ai-behind-faceapp/
  2. Everything we know about the company behind FaceApp, the viral photo app that’s come under scrutiny for keeping your data. (2019, July 18). Business Insider Nederland. https://www.businessinsider.nl/faceapp-privacy-concerns-russian-company-behind-it-2019-7?international=true&r=US#faceapp-reportedly-moved-to-the-skolkovo-innovation-center-which-is-run-by-the-russian-government-in-2018-although-the-firm-said-it-was-not-affiliated-with-the-venture-in-any-way-4
  3. Brewster, T. (2019, July 25). Who’s The Face Behind FaceApp? Meet The Rich Russian Who Built The Wildly Viral App. Forbes. https://www.forbes.com/sites/thomasbrewster/2019/07/24/meet-the-face-behind-faceapp-rich-russian-yaroslav-goncharov/?sh=484a9408255a
  4. Fowler, G. A. (2019, July 17). You downloaded FaceApp. Here’s what you’ve just done to your privacy. Washington Post. https://www.washingtonpost.com/technology/2019/07/17/you-downloaded-faceapp-heres-what-youve-just-done-your-privacy/
  5. Denham, H., & Harwell, D. (2019, July 17). FaceApp went viral with age-defying photos. Now Democratic leaders are warning campaigns to delete the Russian-created app ‘immediately.’ Washington Post. https://www.washingtonpost.com/technology/2019/07/17/faceapp-adds-decades-your-age-fun-popular-russian-owned-app-raises-privacy-concerns/
  6. O’Flaherty, K. (2021, June 30). FaceApp Privacy: What You Need To Know About The Viral Russian App. Forbes. https://www.forbes.com/sites/kateoflahertyuk/2020/06/19/faceapp-privacy-what-you-need-to-know-about-the-viral-russian-app/?sh=1f8dce727219
  7. https://www.faceapp.com
  8. Rosenblatt, K. (2021, December 31). The best memes of 2021: The “yassification” of the internet, feminine urges and feeling cheugy. NBC News. https://www.nbcnews.com/pop-culture/pop-culture-news/best-memes-2021-yassification-internet-feminine-urges-feeling-cheugy-rcna10342
  9. O’Neill, S. (2021, November 24). YassifyBot and “Yassification” Memes, Explained. The New York Times. https://www.nytimes.com/2021/11/24/style/yassify-bot-meme.html
  10. Cresci, E. (2017, April 25). FaceApp apologises for “racist” filter that lightens users’ skintone. The Guardian. https://www.theguardian.com/technology/2017/apr/25/faceapp-apologises-for-racist-filter-which-lightens-users-skintone
  11. Hern, A. (2017, November 27). FaceApp forced to pull “racist” filters that allow “digital blackface.” The Guardian. https://www.theguardian.com/technology/2017/aug/10/faceapp-forced-to-pull-racist-filters-digital-blackface
  12. Prance, S. (2019, July 17). The FaceApp age challenge will show you how old you will look in 60 years time. PopBuzz. https://www.popbuzz.com/internet/viral/faceapp-old-filter-age-challenge/
  13. Manning, A. (2017, May 2). FaceApp’s Gender Swap Is Playing Mind Games With Trans Users. Vocativ. https://www.vocativ.com/424293/faceapp-gender-swap-trans-users/index.html
  14. Associated Press, & McMillan, J. (2019, May 17). In gender-swap photo filters, some trans people see therapy. WHYY. https://whyy.org/articles/in-gender-swap-photo-filters-some-trans-people-see-therapy/