Hey, I froze one of the trained checkpoints to a single file and then applied it on a video that I found on youtube. This is done with OpenCV. If you like to know how to do it, you could check my other article: https://medium.com/towards-data-science/building-a-real-time-object-recognition-app-with-tensorflow-and-opencv-b7a2b4ebdc32. Basically in the article I’m using the webcam as the input source but you can easily modify the code by yourself to use a video as input source instead.