QuickPose

• 

  Pose estimation landmarks overlay

Inspiration

We built an AI powered health app that uses Human Pose Estimation, and built the AI functionality ourselves. We realised that so many people have ideas of using Artificial Intelligence in their apps but there's too much of a tech-barrier, or really expensive to get an AI engineer. With so many ways of using pose estimation in health and fitness apps, we made a SDK so that creators can integrate human pose estimation into their projects easily.

What it does

QuickPose is an iOS SDK that allows developers to integrate human pose estimation into their apps using MediaPipe and BlazePose. We have created features such as : Measuring range of motion of joints, Squat Counter, Jumping Jacks Counter, Push-Up Counter, thumbs up & Thumbs Down capability, and counting fingers shown from hand gestures.

How we built it

We created a wrapper that uses MediaPipe and BlazePose. We trained our own ML algorithms to detect poses of the body for fitness exercises, hand gestures, and do complex calculations of pose landmarks to measure angles between joints.

Challenges we ran into

Training the machine-learning models to accurately predict the poses took a lot of training and a lot of data. The whole team helped by providing videos of doing a range of fitness exercises from different angles and different lighting conditions to ensure that the model was able to learn the pose classifications.

Another challenge was deciding what to build next. There are a whole range of use cases across Health, Fitness, and Gaming, so we want to invite other developers to build the apps using QuickPose.

Accomplishments that we are proud of

We can measure the angles between joints at 2° accuracy, which is more accurate than a goniometer (the tools used by physiotherapists to measure range of motion of joints). As a small team, we made the underlying tech that is used by large AI companies with high numbers of staff and expensive subscriptions for their users to use their products. We are excited to enable developers of all skill levels to increase their skill level and give them the tools to incorporate cutting-edge machine learning into their projects.

What's next for QuickPose

We want to inspire movement through tech. The fitness app industry is worth 19+ billion dollars in 2023 and more and more people do healthy activities through apps. We can't built thousands of apps alone, but by sharing this technology we hope to enable new and experienced developers to build the apps they dreamed of. We are constantly adding to the project and are actively helping people who want to build apps and games that use Artificial Intelligence

Built With

• blazepose
• machine-learning
• mediapipe
• poseestimation
• quickpose
• swift
• visionai
• xcode