With the increasing demand to scan text documents and old books, having a scanner that could automatically detect the orientations of the scanned pages would be greatly beneficial. This paper proposes a fast method to detect orientations based on a support vector machine (SVM), using features developed for each connected component on the scanned page. Results show that the algorithm can achieve an accuracy of 99.2% in orientation detection and 98.2% in script detection for pages scanned at 200 dpi.