The four undergrads met for the first on the Stanford TreeHacks hackathon, grew to become shut pals, and developed an AI-powered app to befriend bodily therapy sufferers be obvious upright posture for their at-home workout routines — all within 36 hours.
Relieve in February, factual sooner than the lockdown, Shachi Champaneri, Lilliana de Souza, Riley Howk and Deepa Marti occurred to take a seat down correct thru from every other on the tournament’s introductory session and nearly straight away made up our minds to have a workforce for the competition.
Together, they created PocketPT, an app that lets users know whether or now not they’re polishing off a bodily therapy exercise with the upright posture and have. It captured two prizes in opposition to a crowded field, and impressed them to continue the exhaust of AI to befriend others.
The app’s AI model makes exhaust of the NVIDIA Jetson Nano developer equipment to detect a user doing the tree pose, a position identified to magnify shoulder muscle strength and toughen steadiness. The Jetson Nano performs image classification so the model can say whether or now not the pose is being done accurately in accordance to 100 photos it became as soon as trained on, which the workforce took of themselves. Then, it provides suggestions to the user, permitting them to clutch if they need to adjust their have.
“It’d be taxing for sufferers to trip to the bodily therapist customarily, both financially and bodily,” acknowledged Howk.
Persevering with workout routines at house is a an crucial fragment of recovery for bodily therapy sufferers, nonetheless doing them incorrectly can if truth be told hinder progress, she outlined.
Bringing the Conception to Lifestyles
In the months main as much as the hackathon, Howk, a rising senior on the College of Alabama, became as soon as interning in Los Angeles, where a yoga studio is form of on every nook. She’d arrived on the competition with the root to manufacture some roughly yoga app, nonetheless it unquestionably wasn’t until the workforce came correct thru the NVIDIA desk on the hackathon’s sponsor horny that they realized the root’s potential to magnify and befriend these in need.
“A demo of the Jetson Nano displayed how the draw can tune bodily motion correct down to the joint,” acknowledged Marti, a rising sophomore at UC Davis. “That’s what sparked the prospective of establishing a bodily therapy app, as opposed to limiting it to yoga.”
None of the workforce people had prior trip working with deep studying and pc imaginative and prescient, so that they confronted the problem of studying implement the model in this kind of short timeframe.
“The NVIDIA mentors were if truth be told handy,” acknowledged Champaneri, a rising senior at UC Davis. “They effect together an tutorial files on exhaust the Nano that gave us the factual footing and clarify to exhaust and implement the root.”
Over the first night time of the hackathon, the workforce took NVIDIA’s Deep Discovering out Institute route on getting started with AI on the Jetson Nano. They’d grasped the basics of deep studying. The next morning, they began hacking and training the model with photos of themselves exhibiting upright versus fallacious exercise poses.
In barely 36 hours for the reason that basis first emerged, PocketPT became as soon as born.
A hit Extra Than Simply Awards
Potentially the most thrilling fragment of the weekend became as soon as checking out the workforce had made it to last pitches, in step with Howk. They presented their project in entrance of a crowd of 500 and later chanced on out that it had obtained the 2 prizes.
The hackathon attracted 197 projects. Competing in opposition to 65 other projects within the Medical Pick up admission to category — a lot of which ancient cloud or other platforms — their project took home the category’s huge prize. It became as soon as also chosen because the “Most fascinating Exercise of Jetson Hack,” amongst 11 other teams that borrowed a Jetson for their projects.
Nevertheless the quartet is having a sight to have more with their app than employ awards.
Thanks to the snappily-paced nature of the hackathon, PocketPT became as soon as simplest ready to utterly implement one pose, with others tranquil within the works. On the different hand, the workforce is dedicated to rising the product and selling their overall mission of establishing bodily therapy without misfortune accessible to all.
While the hackathon took predicament factual sooner than the COVID outbreak within the U.S., the workforce highlighted how their project appears to be the complete more connected now.
“We didn’t even realize we were establishing something that will per chance per chance become the prolonged scoot, which is telemedicine,” acknowledged de Souza, a rising senior at Northwestern College. “We were rising an at-home model of PT, which is terribly great wished factual now. It’s surely price our time to continue engaged on this project.”
Be taught to acquire started on a Jetson project of your bag on the Jetson builders online page.