Page 57 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

pose; otherwise, a buzzer will sound, concurrently highlighting the landmarks in red.
Furthermore, the vision models can be hosted locally to address any privacy-related
concerns.

3 Use Case Requirements 4.1-Healthcare

• REQ01: It is critical to detect all 17 landmarks on the body, based on the input video
stream of the Yoga posture.
• REQ02: It is critical to draw connections between them in white colour.
• REQ03: It is critical to preserve the privacy of user data by processing the same on the
Mobile app itself.
• REQ04: It is critical that model should work with real-time videos of Yoga learners.
• REQ05: It is expected that if the current posture (real time) reaches an accuracy of
threefold above 70% probability, then the colour turns green.
• REQ06: It is of added value to measure the time of holding the posture correctly.
• REQ07: It is of added value to notify the user by playing sound or music.
• REQ08: It is of added value that the mobile app is usable and mobile in perspective.
• REQ09: It is expected that the Mobile app provides a feature for collecting user feedback.
• REQ10: It is expected that the tracking and monitoring will also be a useful feature from
an individual improvement perspective, as well as a research dimension.
• REQ11: It is expected that there should not be any lag, delay, or flickering in providing
the feedback.
• REQ12: It is of added advantage that the Web-based component will help provide model
updates and enhancements.
• REQ13: It is expected that a Security audit should be performed for the mobile & web
application.

4 Sequence Diagram

• Sequence diagram

Figure 2�1: Sequence Diagram of our system

21