How an Indian startup is using a smartphone’s front-facing camera to make fitness fun
Two years ago, when IIT-Bombay graduates Anurag Mundhada and Jayesh Hannurkar were struggling with weight issues during the first wave of the coronavirus pandemic, they thought of using their expertise in sensors and computer vision to create an interactive home workout platform.
“The biggest consumer problem we’re working to solve is motivation,” Mundhada explained the idea behind her Insane AI app. “Super fit models and trainers can be really daunting at times, but if you make the initial onboarding as smooth and seamless as possible, fitness can be fun and is for everyone,” he said. Insane AI, which has been in beta for well over a year and is about to officially launch, is different from other fitness apps, the founders claim.
Unlike Apple Fitness Plus or Nike Training Club, Insane AI uses your smartphone’s front camera to track your movement during a workout, they said. “The camera follows the body and breaks it down into 21 points: your shoulders, elbows, ankles and knees. Once the points are tracked in real-time, we run a video AI layer on top of them, which analyzes the movement and clarifies what exercise is being performed, counts the repetitions, and the system then outputs an assessment of how intense the exercise is.” said Mundhada indianexpress.com during a video call.
“If you squat and your hips are below your knees you get a score of one, and if you’re only doing very flat reps you don’t get the same score,” he added. Mundhada claimed its AI-powered workout tracking technology is 98 percent accurate based on the company’s internal testing data. The intent to use a smartphone as an integral part of your fitness journey and create a new user experience around phone cameras stems from Mundhada and Jayesh’s previous experience with camera sensor technologies. Jayesh was at Sony Japan before co-founding Insane AI, working on LiDAR sensors, a technology that adds deep scanning for better photos and AR in smartphones and has been found mostly in Apple’s high-end iPhones and iPad Pros since 2020. Mundhada, on the other hand, he collaborated with a Bengaluru-based Axogyan AI where he ran a project with Mercedes Benz on human sensing for autonomous vehicles. Mundhada said he and his team were fascinated by the Nintendo Wii and Microsoft Kinect and wanted to offer the ability to track body movement with a phone camera.
Although smartphones have gotten better hardware over the years, depth sensors are still limited to a handful of expensive phones. “Using AI, we’ve enabled the regular RGB camera to track the real-time workout tracking that’s present on all phones to do what the Nintendo Wii or Xbox Kinect can do,” he said when asked how a smartphone camera can track body movements without additional sensors.
To test whether the smartphone camera really tracks body movements and whether an app can be built around home workouts, Mundhada developed the pushup dojo, which went live in the early days of the pandemic. The app was used as a proof of concept to demonstrate that technology and workout tracking could be possible with a phone camera. It was a simple workout app that attracted 2000 users, mostly friends and immediate family of the co-founders.
That gave Mundhada some confidence to revamp the algorithms and UI and perfect the body motion tracking technology to create the Insane AI app in early 2021. This time, the team worked with fitness trainers to create the app and intended a wide user base. Mundhada calls Insane AI a home workout app with a “highly gamified experience.” The app, which can be downloaded from the Google Play Store and Apple App Store, currently requires people to sign up for the waitlist, and starting the last week of October, Insane AI will be available to those currently on the waitlist. To use the app, one must first set up a Cartoon(y) avatar, which will be your virtual identity on the training platform. You can simply place the phone some distance in front of you, start exercising, and the camera will then start tracking your body movement as you exercise.
Every workout or exercise that users perform is counted and they are rewarded with coins that can be redeemed in the form of digital collectibles and coupons, for which the startup has linked up with a few partners. Users can also challenge their friends in both real-time and turn-based fitness battles.
The Insane AI app works on any smartphone launched in 2016 and later. In the coming days, users will see a brand new interface and a variety of features, including many more workouts and exercises.
The AI-powered training technology model developed by Mundhada and his team not only tracks body points, but also tracks users’ fitness levels in five different dimensions such as strength, cardio, upper body and lower body. Mundhada said they took a lot of input from fitness experts to develop the training plans and decide on different fitness levels, while the company’s in-house data science team worked to create a fitness model that could be customized and customized based on the user can be personalised.
The insane AI app offers 35 different exercises and 150 workouts, 4 of which are free to start with. According to Mundhada, Insane AI seems to be widespread at 7:30 am, and another peak comes in the evening around 6 pm. On average, users spend nine minutes on the app, up from 5 minutes in the early days of Insane AI.
However, power users spend almost an hour on the platform. “Insane AI isn’t meant to completely replace a personal trainer,” Mundhada said, adding the app is “designed to provide an easy path to fitness learning and progression.” “With cases like extreme obesity, it’s always wise to have someone in the loop, and not just for learning, but for injury prevention. Someone who can help you recognize you lifting weights.” Mundhada understands that Insane AI aims to provide a rewarding and fun experience and to complement your journey along the way. The home workout app promises not to collect users’ images and videos, only collecting anonymized movement data itself to improve the AI engine that performs movement tracking. For monetization, Mundhada says they are opting for a freemium model, allowing users to experience the app for free, although certain features like advanced challenges and redeeming rewards require a subscription. At the moment, those who sign up for the app get a free subscription for up to six months. Mundhada and the team have yet to figure out what a subscription would cost. Last year, the startup raised $900,000 in seed funding. The round was led by early-stage venture capital fund pi Ventures, Anupam Mittal, Sameer Pitalwalla and other esports, fitness and tech angels. In the long term, Mundhada would like to make Insane AI a fitness metaverse and work to create a foundation for an immersive world that has already begun.