Test Data Set



Next: Unfiltered Results Up: Results and Analysis Previous: Basic Processing Model

Test Data Set

The test data set consists in 439 pages that were taken from ISRI's Sample 2 Document Database. The set of pages have the following characteristics:

The pages were processed by eight OCR devices (see Table 4.1). The median OCR accuracy was computed for each page from the results of these eight devices and that was the accuracy value used to label the pages as ``Good'' or ``Bad''.