Skip to content

Personal Dataset

Latest
Compare
Choose a tag to compare
@vitostamatti vitostamatti released this 23 Aug 16:49
· 41 commits to main since this release

Release components

  • Landmarks training datasets with pose data augmentation using Yolo, Mediapipe and Movenet pose models
  • Trained pose classifiers models with differents landmarks datasets (Yolo, Mediapipe and Movenet) using sklearn Random Forest estimator.

Release final observations

Raw training image dataset:

- Fall: 844 images
- No Fall: 645 images

Yolo training CSV's observations:

- Fall augmented: 4180 / 4220
- No Fall augmented: 3222 / 3225

Movenet training CSV's observations:

- Fall augmented: 4220 / 4220
- No Fall augmented: 3225 / 3225

Mediapipe training CSV's observations:

- Fall augmented: 3777 / 4220
- No Fall augmented: 2930 / 3225