Need to look into a better way to use custom trained models. Right now, we're using FritzAI and it is not free to use custom models. If we want to enhance and detect more objects, we have to change API.
Refer to https://towardsdatascience.com/real-time-mobile-video-object-detection-using-tensorflow-a75fa0c5859d