Imagine running late for a flight, only to be flagged by a facial recognition (FR) system at the airport. You’ve done nothing wrong, but the system mistook you for someone else who isn’t allowed to fly. Unfortunately, by the time everything is sorted, you’ve missed the plane.
FR systems are increasingly showing up in airports, stadiums, at borders and more. Their rapid expansion is increasing the need for ethically-sourced synthetic datasets that improve how FR algorithms recognize race and gender and mitigate the model performing better for specific demographics than others.
At SMU’s Intelligent Systems and Bias Examination Lab (ISaBEL), researcher Corey Clark and his team are generating facial image datasets from text descriptions with the university’s NVIDIA DGX SuperPOD, a high-performance computing platform specifically designed for AI.
ISaBEL was founded as a partnership between the foremost commercial developers of AI systems and the leading scientists, faculty, and students at SMU. The lab’s goal is to understand and mitigate bias in AI systems using the latest research, standards, and other peer-reviewed scientific studies. Pangiam, a global leader in artificial intelligence, is ISaBEL’s first industry partner.
Clark and his group are currently producing datasets containing millions of facial images that include underrepresented racial groups. This methodology gives FR systems a greater chance of being more fair and accurately balanced.
“There are constraints in trying to create a real-world based dataset to train any artificial intelligence model,” said Clark, assistant professor of computer science in the Lyle School of Engineering and deputy director for research at SMU Guildhall. “To ethically source it you must solve challenges like consent, fairness, and legal compliance. Synthetic data, generated by the SuperPOD, removes those obstacles.”
Existing FR technology has struggled to match the same face with different angles and poses. Using an existing stable diffusion model (an open-source AI algorithm that anyone can use), the SMU researchers have generated large, diverse datasets with pose variations.
Their customization of the stable diffusion model is unique due to the sheer magnitude of images created – millions so far – and the special tuning of the model to specifically process facial recognition.
“Facial recognition is here and not going away,” Clark said. “The demand for these larger training datasets is crucial for improving FR systems so they provide equitable results. Through our methodology and use of the of the SuperPOD, we’re generating datasets not previously easy to obtain, and doing so quickly and ethically.”
In 2021, SMU announced its collaboration with NVIDIA, a trailblazer in the field of accelerated computing, through the University’s acquisition of an NVIDIA DGX SuperPOD, which expanded SMU’s supercomputer memory capacity and led to a 25-fold increase in the speed and efficiency of AI and machine learning.
Clark stressed that the massive number of images created for their datasets would not be possible without the SuperPOD, and its capabilities will have a significant role to play in further FR development. Moving forward, Clark and his team plan to create one of the largest balanced facial recognition data sets for research use.
By addressing fairness and bias issues found in FR technology, Clark and his colleagues also plan to create a bias certification process that could evaluate existing companies’ AI and be used to develop future models specified to need.