YogaFix is an application designed to help users practice yoga poses correctly by providing real-time feedback and guidance. It uses computer vision to detect body landmarks and a combination of an expert system and a deep learning model [CNN] to analyze and correct yoga postures.
- Real-time Pose Detection: Utilizes MediaPipe Pose to detect 33 key body landmarks in real-time from a webcam feed.
- Expert System for Pose Correction:
- Defines correct joint angles and acceptable ranges for various yoga asanas in a flexible JSON configuration file (
exercise_config.json). - Provides specific corrective feedback if joint angles are outside the defined ranges.
- Visual feedback: Draws lines on the user's body, colored green for correct angles and red for incorrect ones.
- Defines correct joint angles and acceptable ranges for various yoga asanas in a flexible JSON configuration file (
- Guided Asana Sequence:
- Presents the name and description of the current asana.
- Shows a reference image of the target pose.
- Monitors the duration for which a pose is held correctly.
- Text-to-Speech (TTS) Feedback:
- Provides audio cues for pose names, descriptions, and corrective feedback.
- Runs in a separate thread to avoid UI freezes.
- CNN based Pose Classification:
- Aims to automatically identify the yoga pose being performed by the user using a Deep Learning model trained on landmark data.
- Once classified, the expert system takes over for detailed correction of that specific pose.