The method detailed herein relates to a classifier for classifying a video as an advertisement or program without using a reference database of frames. The method can include classifying videos with a single-frame classifier that analyzes a single frame to determine whether the frame is an advertisement or a program. Furthermore, the method includes classifying videos with a sequence classifier that analyzes sequences of frames to determine whether the frame is an advertisement or a program. In some embodiments, the method relates to classifying predefined amounts of frames as advertisements or programs to determine whether a video is an advertisement or a program. In some embodiments, the method relates to classifying frames of a reference database of an advertisement matching system in order to adjust various analysis weights so that the performance of the advertisement matching system does not fall due to locally inserted advertisements that may not be part of the reference database.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Mehta, Devraj and Solanki, Ravi, "SYSTEMS AND METHODS FOR DEEP LEARNING TELEVISION (TV) ADVERTISEMENT DETECTION", Technical Disclosure Commons, (September 12, 2018)