In the spines directory are sample images of books which were photographed from above. To be processed, we need to separate the pages into separate images, with the left side page (verso) in one image, and the right side page (recto) in another image.
While we work with hundreds of thousands of images, perhaps 10-30% need to be split. While splitting can be done programmatically already, we need to find a way to detect the location on the original image on which we should split. To that end, we've provided a gold data set which was manually classified:
In the spines directory are sample images of books which were photographed from above. To be processed, we need to separate the pages into separate images, with the left side page (verso) in one image, and the right side page (recto) in another image.
While we work with hundreds of thousands of images, perhaps 10-30% need to be split. While splitting can be done programmatically already, we need to find a way to detect the location on the original image on which we should split. To that end, we've provided a gold data set which was manually classified: