The company that developed the Snapchat application launched an initiative to redesign its basic camera technology to make it more capable of capturing a wide range of skin tones. About 5 billion photos are captured using the Snapchat camera every day, and these images form the starting point for people who see themselves and their friends. And their world through it.
Historically, the chemical processes behind the film’s development used light skin as its chemical baseline, a legacy that continues today, but the camera is actually racist, according to TechNews Arabia.
Film cameras are getting better at exposing the darker colors, but not as part of a concerted effort to make things more fair to people.
Instead, it was complaints from chocolatiers and photographers shooting other dark subjects that drove the industry to do better.
The early days of digital photography were similar, as some corporate web cameras like Microsoft promised the ability to detect faces, but had a hard time doing so with dark-skinned people.
However, technology has made strides in recent years that will aid this effort, including high dynamic range and the ability to combine multiple shots to create a single image.
For Snapchat, the all-in-one camera effort is broader than just capturing dark skin as well as fair skin, and that means identifying and removing biased assumptions when automatically adjusting people’s looks.
Snap still wants people to be flexible, but it wants to make real, high-quality image the starting point, and then put the controls in one’s hands.
The company-wide effort began with a presentation by Saint Brieux to top executives last summer in the wake of George Floyd’s protests.
Snap is working with several famous photography directors from the movie industry to learn the techniques they use to capture the best shots of dark-skinned actors.
Among the projects under development or testing:
• Developing techniques to adjust images after capturing them, such as: correcting the brightness and exposure in order to create a more balanced image.
• Improving the selfie cameras ability to capture low light by making adjustments to the front flash, so that if someone is taking a selfie in a dark room, the screen uses the right type in order to properly highlight the color of their skin.
• Machine learning systems and how these systems are improved, along with a focus on making everyone’s image quality above a certain threshold in order to obtain a more equitable result.
Snap acknowledges that its track record is far from perfect, and she said: We are well aware of our past mistakes and apply what we have learned to all our efforts to build more comprehensive design processes, systems and products.
Snap is working on a variety of efforts that take longer to bring them to market, and one part is to expand the overall camera effort to other groups, such as ensuring that those who wear glasses or other aids can use the company’s filters fully.
Snap is also looking at how to enable third-party developers and partners to take advantage of the tools you build internally.