Monday, March 4, 2024
HomeUncategorizedFitness AI: How Synthetic Data Powers Better Workouts

Fitness AI: How Synthetic Data Powers Better Workouts

Can’t you attend Transform 2022? Check out all the summit sessions in our on-demand library now! look here.

AI-driven app developed using synthetic data can analyze the details of how people move ?

Home fitness apps are all the rage during the COVID-19 pandemic. From January 2020 to November 2020, approximately 2.5 billion health and fitness apps were downloaded worldwide. This trend continues and shows no signs of slowing down, with new data forecasting growth from $10 million in 2022 to $23 million by 2026.

As more and more people use fitness apps for training and tracking With the development and performance of fitness applications, more and more fitness applications AI-based workout analysis, combining techniques such as computer vision, human pose estimation, and natural language processing techniques, is increasingly using AI to power its products.

Founded in 2018, Tel Aviv-based Datagen claims to provide “high-performance synthetic data, focused on data for human-centric computer vision applications.”


MetaBeat 2022

MetaBeat will be on October 4th at San Francisco brings together thought leaders to provide guidance on how Metaverse technology is changing the way all industries communicate and do business, CA.

Register here

The company just announced a new domain name Smart Fitness on its self-service visual synthetic data platform to help AI developers generate the data they need to analyze people’s workouts and train smart fitness devices to “see”.

“At Datagen, our focus is on helping computer vision teams and accelerating their development of human-centered computer vision tasks,” Datagen CEO Ofir Zuk told VentureBeat. “Almost every use case we see in artificial intelligence is related to humans. We are specifically trying to address and help understand how humans are interconnected and how they interact with their surroundings. We call it humans in this context. ”

image gif showing datagens body key point analysis for workouts Synthetic visual data representing the fitness environment

The Smart Fitness platform provides 3D annotated synthetic visual data in the form of videos and images. This visual data accurately represents the fitness environment, high-level motion, and human-object interactions for tasks related to body keypoint estimation, posture analysis, posture analysis, repetition counting, object recognition, and more.

Additionally, teams can use the solution to generate full-body motion data to iterate their models and rapidly improve their performance. For example, in the case of pose estimation analysis, an advantage offered by the Smart Fitness platform is the ability to quickly simulate different types of cameras to capture a variety of differentiated motion synthesis data.

data gen Smart Fitness platform post tracking gif of person doing lunge

Source: Datagen

Challenge of training AI fitness

Pose Estimation, which is a computer vision technique that uses images to help determine the position and orientation of a person, is one of the unique solutions that artificial intelligence has to offer. For example, it can be used for avatar animation of artificial reality, as well as markerless motion capture and worker pose analysis.

To properly analyze the pose, it is necessary to capture several images of the environment the human actor interacts with. These images are then processed by a trained convolutional neural network to predict where the human actor’s joints are in the image. AI-based fitness apps often use the device’s camera to record video at up to 720p and 60fps to capture more frames during a workout.

The problem is that computer vision engineers need a lot of visual data to train AI for fitness analysis when using techniques like pose estimation. Data involving humans practicing in various forms and interacting with multiple objects is complex. The data must also have high variance and be sufficiently diverse to avoid bias. It is nearly impossible to collect accurate data covering such a wide range. On top of that, manual annotation is slow, prone to human error, and expensive.

While the accuracy of 2D pose estimation has reached acceptable levels, 3D pose estimation is lacking in generating accurate model data. This is especially true for inference from a single image and no depth information. Some methods utilize multiple cameras pointed at a person to capture information from depth sensors for better predictions.

However, part of the problem with 3D pose estimation is the lack of large annotated datasets of people in open environments. For example, large datasets for 3D pose estimation (eg Human3.6M) are captured entirely indoors to remove visual noise.

We are working to create new datasets with more data on environmental conditions, clothing types, articulation and other influencing factors.

data gen Smart Fitness platform post tracking gif of person doing lunge Integrated Data Solutions

To overcome these issues, the tech industry now widely uses synthetic data, which is artificially generated data that closely mimics operations or production Data, used to train and test artificial intelligence systems. Synthetic data provides several significant benefits: It minimizes restrictions associated with using regulated or sensitive data; can be used to customize data to match real data does not allow conditions; and it allows for large training datasets without manually labeling the data.

As reported by Datagen, the use of synthetic data reduces production time, eliminates privacy concerns, provides reduced bias, annotation and labelling errors, and improves predictive modeling. Another advantage of synthetic data is the ability to easily simulate different camera types while generating data for use cases such as pose estimation.

data gen Smart Fitness platform post tracking gif of person doing lunge Practice demo made easy image gif showing datagens body key point analysis for workouts

With Datagen’s smart fitness platform, organizations can create tens of thousands of unique identities to perform in different environments and conditions Various workouts – in a fraction of the time.

“With powerful synthetic data, teams can generate all the data they need with specific parameters in a matter of hours,” Zuk said. “Not only does this help retrain the network and machine learning model, it also allows you to fine-tune it in no time.”

data gen Smart Fitness platform post tracking gif of person doing lungedata gen Smart Fitness platform post tracking gif of person doing lungedata gen Smart Fitness platform post tracking gif of person doing lungeSource: Datagen

In addition, he explained that the Smart Fitness platform optimizes your ability to capture millions of massive visual motion data, eliminating duplication Sexual Burden captures each element in person.

“Through our constantly updated library of avatar identities and motion types, we provide detailed pose information, such as the position of body joints and bones, which facilitates analysis of complex augmentations Details of the AI ​​system,” he said. “Adding such visual capabilities to fitness apps and devices can dramatically improve the way we look at fitness, enabling organizations to deliver better in-person and online services.”

Source: Datagen data gen Smart Fitness platform post tracking gif of person doing lunge

Fitness AI and Enterprise Synthetic Data
image gif showing datagens body key point analysis for workouts

According to Gartner Distinguished Vice President Analyst Arun Chandrasekaran, synthetic data is by far a “higher level of enterprise adoption.” Low Emerging Technologies”.

However, he said that for use cases where data anonymity must be guaranteed or privacy must be preserved, such as medical data, will see increased adoption; increasing real data, Especially when data collection is expensive; there is a need to balance class distributions in existing training data, such as population data, with emerging AI use cases with limited real-world data available.

Several of these use cases are key to Datagen’s value proposition. When it comes to enhancing the functionality of a smart fitness device or app, “of particular interest is the ability to improve data quality, broad scenario coverage and privacy protection during the machine learning training phase,” he said.

Zuk admits that bringing AI and synthetic data, or even digital technology in general, into fitness is still in its early stages.

“They’re very unpassive and very lean in terms of capabilities,” he said. “I would say adding these visual features to these fitness apps, especially as people are doing more exercise at home, will definitely improve things significantly. We’re clearly seeing an increase in demand and we can imagine people being able to What to do with our data.”

The Mission of VentureBeat is a digital city square for technology decision makers to gain knowledge about transformative enterprise technologies and transactions. Learn more about membership.



Please enter your comment!
Please enter your name here


Featured NEWS