🇺🇸 Michelle Hua

The prize and the project

The George D. Yancopoulos Innovator Award is the top individual honour at the Regeneron International Science and Engineering Fair, and in 2021 it carried $75,000. The Society for Science named Hua, 16, of Troy, Michigan, as the winner in its announcement of 20 and 21 May.

Her project was an artificial-intelligence algorithm for human action recognition. Rather than working from full-colour video, it used a deep-learning framework to analyse silhouettes — the outline of a moving body, stripped of colour, texture and background. According to the organiser's summary, the method outperformed existing state-of-the-art approaches.

Why silhouettes are the interesting choice

Action recognition is a standing problem in computer vision: telling walking from falling, reaching from throwing, from pixels alone. Most systems attack it with as much visual information as they can get, which makes them heavy to run and dependent on lighting, clothing and camera quality.

A silhouette discards nearly all of that. What remains is shape and how it changes, which is cheaper to process and far less revealing of who the person is. The organiser's description credits the approach with beating the previous best results; it does not publish the benchmark datasets, accuracy figures or code, so the margin of that improvement is not on the public record.

There is a second advantage in the same choice. A system that never records a face or a garment is far easier to place in a hospital corridor, a care home or a factory floor, where the useful question is whether someone has fallen rather than who they are. Privacy, in that design, is a by-product of the method rather than a policy bolted on afterwards.

The fair that had to move online

Regeneron ISEF 2021 was the first virtual edition in the competition's history of more than seventy years. It still assembled over 1,800 finalists, representing 49 US states and 64 countries, competing for more than $5 million in awards.

Maya Ajmera, president and chief executive of the Society for Science, framed the year in those terms: "Every single Regeneron ISEF finalist has persevered in their pursuit of science in the face of the myriad challenges of the COVID-19 pandemic." Judging, presentation and the awards ceremony all ran over video.

Who else stood at the top that year

Below the Yancopoulos Award sat two Regeneron Young Scientist Awards of $50,000 each, taken by Catherine Kim, 18, of Jericho, New York, for a machine-learning model predicting adverse drug reactions, and Daniel Shen, 17, of Cary, North Carolina. John Benedict Estrada, 16, of Fresno, California, took the $50,000 Gordon E. Moore Award for Positive Outcomes for Future Generations.

The $10,000 special awards went to Arya Tschand of New Jersey for irrigation AI and Franklin Wang of California, whose machine-learning search of public telescope data turned up six previously unknown near-Earth asteroids. Hua's award was larger than any of them.

What is documented — and what is not

The firm record is the organiser's own: the award, the amount, the age, the home town and a one-paragraph description of the method. That is strong evidence for the result and thin evidence for everything else.

No source gathered here reports where Hua studied afterwards, whether the algorithm was published in a journal or released as code, or what she is working on now. Those are gaps in the reporting, not conclusions about her. Readers searching for a current affiliation will not find one documented in this record.