Trying to implement multi-cam object detection

Thanks for your work! I could improve FPS from 3 to 9 (Yolov3). Now, I would like to develop a multi-cam based object detection. Currently, when I try to spawn out new threads for each camera, I keep getting tensorflow session errors. Could you please guide me on how to set-up multi cameras for object det?  