We analyzed five hours of data from five diverse films (an hour from each film). Three English films: Sense and Sensibility, Pulp Fiction, Four Weddings and a Funeral. Two foreign films: Farewell my Concubine (Chinese), Bombay (Hindi). We used the foreign language films to test the impact of language on the seemingly low-level task of segmentation. We present our computable scene detection results below:

The green bars indicate precision while the blue bars indicate recall. The best case results are for Bombay with 88% precision and 72% recall. The results for Sense and Sensibility are also very good with nearly 80% precision and recall.