The top-down view problem is what I was facing when I was trying to use a detection model for a project with quadcopters. For this I had to train my object dete

The top-down view problem is what I was facing when I was trying to use a detection model for a project with quadcopters. For this I had to train my object detector with the data we collected from that view. YOLO was used for this work. Yes, replacing the fine-tuned model with compiled graph should work. I tried to keep it simple and intuitive to change the model so as to allow the use of any other custom models.