/autozoom

Primary LanguageC++

Auto zoom to 70% (front camera) Auto zoom to 46% (Back camera) Auto zoom to 40% (Back camera)
video1.mp4
video2.mp4
video3.mp4

App Flow

  • Turn on the camera services using camera package
  • Detect face using google_mlkit_face_detection package
  • Pass detected faces to the zoom calculator continuously
  • Zoom is based on each previous frame to increase accuracy
  • After each frame zoom level is getting calculated and passed to a camera controller.
  • Update the streaming screen using camera controller
Approach 1 Approach 2
Zoom value is based on each previous detected frame Zoom value is based on the first detected frame
It calculates zoom value by comparing regionHeight and regionWidth with screenHeight and screenWidth It calculates zoom value by a mathematical formula which involves, regionHeight, regionWidth, screenHeight, screenWidth
Zoom level is adjusted seamlessly even when the camera is moved forward or backward. Which ensures real time auto zoom adjustment feature. Zoom level is adjusted with very less accuracy when the camera is moved. Real time auto zoom adjustment feature doesn't work well here.
This approach is also faster but not as of approach 2 because it involves continuous comparison for each new frame. This approach is very faster in time because it involves mathematical formula
Accuracy is high for both nearer and distant objects. Less accurate for nearer objects.

I preferred approach 1 because it gives real time features with more accuracy.