[LEGACY] Data Collection Protocol

[Timothyrichpt]

Collection

Instruct the patient

Stand still with your arms crossed for about 3 seconds. Then, simulate a fall and recovery for about 3 seconds. Repeat this loop 5 times.

Conditions for accurate detections

• Not too close: Patient needs to be fully visible.
• Not too far: Patient needs to occupy at least 60% of the screen height.
• Not too fast: Patient needs to be moving at a reasonable pace to simulate a normal assessment environment.
• Fixed Camera Position: App user needs to be static to simulate a normal assessment environment.

Storage

• PNG image of the patient
• CSV file of the image metadata and the detection
  • Metadata: imageHeight, imageWidth, landmarkRatio, classification<0,1>
  • Detection: [33,5] Pose Output

---

Validation

As the data collector, you can expect to press the button about every 3 seconds.

Input: True

• PNG and CSV are moved within the users' private storage folder raw/ -> /valid

Input: False

• PNG and CSV are moved within the users' private storage folder raw/ -> /invalid

[charlieforward9]

They need to be automatically uploaded, I am creating that pipeline now with:
#45
#47
#48
#49

[charlieforward9]

@Timothyrichpt All of these issues are closed. You are now able to collect and validate data with the app. Its not very user friendly yet, so Ill call you tomorrow to explain it.