A computer implemented method detects scene boundaries in videos by first extracting feature vectors from videos of different genres. The feature vectors are then classified as scene boundaries using a support vector machine. The support vector machine is trained to be independent of the different genres of the videos.