In conversation with TECNO: Leading the way to end skin tone bias in smartphone imaging

Smartphone camera technology is now incredibly advanced, with increasingly capable hardware complemented by fantastic AI-driven software that takes phone photography to the next level. However, despite all the AI smarts, or maybe because of it, since many AI systems are “learning” from obsolete data sets for camera imaging based on fairer skin types, one area where smartphone cameras suffer is skin tone accuracy and representation.

TECNO hopes to bring more inclusivity to phone photography and to end skin tone bias with its innovative Universal Tone technology and the launch of the #ToneProud campaign. Android Authority had the opportunity to speak to Elva Zhou, Director of Image Quality Testing and Assessment Department (IQA), TECNO Image Quality Testing Lab, about the exciting new feature and how it is a potential game-changer globally, especially in emerging smartphone markets.

Universal Tone is TECNO’s answer to widespread skin tone bias

Skin tone bias is a well-documented phenomenon that has exploded in relevance as smartphones have put increasingly capable cameras into the hands of billions of users worldwide. We all love to take pictures of friends and family and we do so in a staggering variety of ways and in many different environments. The stakes are high: we humans are amazingly perceptive when it comes to spotting details in other people’s appearances, including the tone of their skin. Even minute differences between what we perceive with our eyes and what cameras capture can result in images feeling “off” – even when we can’t quite put a finger on the issue.

Inaccurate skin tone representation may seem like a fairly minor issue to some, but hundreds of millions of smartphone users are affected by it, some even without realizing it. And the problem goes back many years — or decades if you consider that film photography suffered from this issue as well. People with darker or simply less-common skin tones often look unflattering in pictures. Social media and forums are rife with examples of users expressing confusion and dissatisfaction over the issue.

The industry has made some efforts to address this phenomenon. Google for instance has made better skin tone representation one of the selling points of its Pixel phone lineup. However, skin tone bias remains prevalent.

TECNO developed Universal Tone to combat the difficulty faced by users when it comes to smartphone photography not accurately reflecting their skin tones, a problem that is particularly prevalent in emerging markets. Whereas giants like Apple, Samsung and Google focus on developed markets, TECNO is deeply familiar with a more diverse set of markets, including countries throughout Africa, the Middle East, and Southeast Asia.

First unveiled by TECNO in 2023, Universal Tone is the company’s multi-skin-tone imaging technology powered by AI. At its core lies the industry’s largest and most accurate skin tone database, the only one to offer 268 skin tone patches.

Zhou says that Universal Tone was created to accurately and properly represent the color and texture of every skin tone in smartphone photography. She points out that this is not innovation for the sake of it or just following industry trends. Accurate skin representation is an issue that consumers in TECNO’s core markets care deeply about.

Universal Tone combines a multi-skin tone restoration engine, a local tuning engine, and a computational portrait engine to create a unified system that delivers true representations of diverse skin tones in images. According to Zhou, TECNO has worked with color science academics at leading global universities to develop this technology.

Unsurprisingly, one of the main factors behind TECNO’s decision to build this technology was consumer demand. Its market research showed that skin tone inaccuracy was the most common complaint when it comes to imaging.

The fulminant rise of AI has further compounded the problem. That’s because depending on where AI tech is developed, skin color processing algorithms can be biased towards certain characteristics while ignoring the diversity of people in emerging markets. When AI algorithms are trained on databases of mostly fair skin tones, they will inevitably show a bias towards fairer complexions. 

As an example, Zhou says that in Africa typical smartphones are prone to producing reddish or overly darkened skin tones. Meanwhile in South and Southeast Asia, varying lighting environments can lead to color misrepresentation, while in Eastern Europe even fair skin tones are prone to overexposure.

Not only is TECNO well-equipped to serve users in these regions, but it’s also the only smartphone maker that has put skin color diversity at its core of its strategy, says the executive.

Whereas other smartphone brands tend to use frequently seen skin tones in North America, Western Europe, and East Asia as their benchmark for skin color recognition, TECNO flipped the script and started its imaging tech development with darker and underrepresented skin tones as the foundation. 

Consumer research, data, and technology advancements, along with years of research and development for emerging markets, put TECNO in a unique position to develop this inclusive technology.

According to Zhou, TECNO went beyond just trying to adapt existing technology — the company set to completely reimagine camera tech with diversity in mind. The large, inclusive skin tone database was just the beginning, with advanced skin tone restoration and calibration AI algorithms helping to further refine the results. Not only does this ensure that the system is inclusive right from the start, but it also solves the question of who the brand is serving with its smartphones.

Unlike the systems built by other phone makers, Universal Tone is a complete skin color evaluation system that considers factors like skin color recognition, classification, color grading, calibration, and enhancement. Of course, skin tone itself is just a part of the equation. The environment and shooting conditions also need to be taken into account – even identical skin tones will look vastly different under fluorescent light compared to natural light for instance. 

“One of the core aspects of the Universal Tone technology is skin color evaluation and restoration – this capability is achieved by AI algorithms,” says Zhou. However, she points out that an unbiased and comprehensive database of reference images remains crucial in order to achieve good results.

This is where TECNO sets itself apart by putting together its database of 268 skin tone patches to help fine-tune the algorithms. This industry-first achievement ensures the coverage of skin tone characteristics found in emerging markets.

TECNO has asked its local teams from around the world to engage in localized testing for typical scenarios. The evaluation process included tests for skin tone representation under various color temperatures, lighting conditions, and other factors. 

The executive also called out some of the weaknesses of current technologies, such as limited or biased databases, or the limited ability to adapt to difficult lighting conditions or unusual skin tones. 

So far, TECNO has focused on key development areas like the skin tone database and measurement, skin color detection and correction, and skin color AI algorithms. However, development progress is full steam ahead. In the next three years, TECNO plans to incorporate generative AI in skin imaging, which will significantly improve the diversity and accuracy of skin color restoration. Expanding the skin tone database beyond the current 268 patches is also a priority.

Of course, the biggest goal is to bring Universal Tone to as many TECNO smartphones as possible. While the system is currently available on mid-to-high-end devices like the PHANTOM V Fold2 5G, PHANTOM V Flip2 5G, and the CAMON 30 Series, TECNO intends to expand the technology to all of its product lines going forward.

Zhou says that over the next decade, we can expect more progress towards realistic and accurate imaging when it comes to skin color representation. Better camera specs, computational photography techniques, and continuously improving AI algorithms, will all play a role.

TECNO launched the #ToneProud campaign on November 4, 2024, to raise awareness about and tackle the problem of skin tone bias in modern smartphone photography. 

The campaign kicks off with a moving film, Tone Proud, featuring celebrity campaign supporters, including Indonesian-born singer-songwriter Anggun, Saudi Arabian filmmaker and actress Fatima Al-Banawi, Nigerian singer-songwriter Johnny Drille, and Polish actress Ewa Kępys, who share their experiences of skin tone misrepresentation and advocate for accurate representation.

TECNO also opened access to its unique 268 skin tone patch database, so anyone can discover their actual skin tone code. The company hopes this will encourage more people to speak out about this pervasive issue. TECNO also plans to launch other activities on Instagram to allow people to share their own multi-skin tone photography experiences. As Zhou summarizes, “skin color is not defined by race nor simplistically divided into black and white. Skin tone is diverse and varied, and everyone should feel proud and confident in their own skin tone.”