SB05a-21 Power Fitness, A Human Pose Estimation Based Indoor Fitness Monitoring and Grading System

Subscribe
Notify of
4 Comments
Inline Feedbacks
View all comments
MA, Chit Ming
May 13, 2022 2:38 pm

Great work!!!!

LAI, Chi Wan
May 13, 2022 2:43 pm
Reply to  MA, Chit Ming

Thank you!

HUI, Chi To
May 13, 2022 2:20 pm

I saw on the poster that you use CNN, have you built a model of your own? How many datasets did you use for each exercise to train the model?

LI, Chun Ming
May 13, 2022 2:33 pm
Reply to  HUI, Chi To

Sure, we did build and apply a neural-network based Pose Classifier using Python in Power Fitness. This Pose Classifier is mainly used in Yoga Pose Detection and Grading. For each pose, we used a total of 120 photos (For Left: 60, Right: 60). The photos are all acquired and captured by professional Yoga coaches on the Internet.
You can build a model by yourself using the tutorial below~
Notebook FYI: https://www.tensorflow.org/lite/tutorials/pose_classification