Posture detection model

Hello,
I’ve been working on project titled ‘posture detection system’. The aim is to develop a model that tells whether the human pose is correct or not. The detection is not real time i.e. we will be training our model on a dataset having ‘good’ and ‘bad’ postures. We are quite puzzled about the roadmap we should follow (from gathering the dataset to choosing and training the model).

Advices are most welcome…

What machine learning courses have you attended?

Take it with a grain of salt from someone who only took the course and had no actual experience.

For collecting data:

  • image - easier to gather and will need less training data
  • sensors on a body - will most likely produce better results
  • combo of both

Annotation:

  • split bad/good postures into separate directories

Training:

  • from scratch if data comes from sensors
  • use smth like YoloNAS and add a layer to classify good/bad postures
1 Like