How to use TensorFlow for comparing key points?

Maria Salim

unread,

Mar 29, 2021, 11:17:19 AM3/29/21

to ML on Mobile OS Working Group

I am working on a fitness android application that would detect if a person has exercised correctly or not.

The app will record the person exercising, then compare it with the models, and display if the person has exercised correctly or not.

I have googled a lot about it; I came across many blogs/articles/posts that helped identify key points, but I couldn't find anything related to comparing the key points. I couldn't even find anything about comparing key points in images. All I could find was how TensorFlow detects key points or recognizes certain poses. I also tried https://teachablemachine.withgoogle.com/train, but the pose estimation only has a .js export for the web.

How will the machine be trained with the key points? How will the key points be compared? Can you suggest any other API that would be easier to understand and use in this regard?

I am new to TensorFlow and would highly appreciate it if someone can help me.

Ben B

unread,

Apr 14, 2021, 9:45:45 AM4/14/21

to ML on Mobile OS Working Group

You're probably looking for pose classification? Here is a nice place to start: https://developers.google.com/ml-kit/vision/pose-detection/classifying-poses

Maria Salim

unread,

Apr 15, 2021, 6:30:01 AM4/15/21

to ML on Mobile OS Working Group

Thank you, I'll check the link

Reply all

Reply to author

Forward