This code uses pre-traind efficient net model on ImageNet dataset to create a video fingerprint These codes are in video level: We create a fingerprint from each 30 frame and make a fingerprint from each frame(picture). The average of the whole fingerprint will be computed and will be used to compare videos together. The whole goal is to detect and find a copy video if there is one in the dataset with use of fingerprints. The ScaNN has been used for a searching algorithm. This project can be used in websies like youtube in order to not to a copy video beeing uploaded by a user! This level of comparing is really not a specefic one while it need to be. but We are working on the next level to...! The next level where the similar videos has been found is to compare videos frame by frame even if the lenght and the place of the copy parts are not the same!