Safely Back to Work
Fig . 14 : People obscuring the view of those standing behind them
People wearing clothes of color similar to the background may be imperceptible by such a system .
Fig . 15 : People wearing clothes matching with the color of wall
Video Analytics for Mask Detection
For using the video analytics method for monitoring facial mask compliance within premises , it ’ s very important that cameras are placed in a way so as to capture almost-frontal faces of people . Most of the algorithms for mask detection work well only up to a certain angle of the face with respect to the camera . A good way to ensure proper facial orientation would be to place cameras directly opposite to doorways so that people face the camera while entering into the room . Along with proper placement , performance of the system will also depend on ambient lighting conditions , camera resolution , etc .
Mask detection algorithms are mostly deep learning models , whose performance depends highly on the examples that the model is trained on . For an accurate classification , it ’ s crucial to train the model on a good number of both masked and unmasked human faces . It is also important to
IIC Journal of Innovation - 33 -