Latest OCR evaluation is done by evaluating each "box" against the gold data. However, this will not accurately evaluate the OCR results from the entire still image, simply because the gold annotation is done on frame-by-frame basis, not word-by-word.
Latest OCR evaluation is done by evaluating each "box" against the gold data. However, this will not accurately evaluate the OCR results from the entire still image, simply because the gold annotation is done on frame-by-frame basis, not word-by-word.