Consolidation of bounding boxes generated from text type pattern localizer

A document copy machine comprising:

• a scanning element;
• logic configured to transform bounding boxes generated by one or both of a region proposal network and a region of interest network by:
• a) determining a first set of the bounding boxes exhibiting dense spacing;
• b) determining a second set of the bounding boxes exhibiting sparse spacing;
• c) executing a first transformation to merge at least some of the bounding boxes exhibiting dense spacing to generate refined bounding boxes, the first transformation comprising:
• detecting that a smaller bounding box of the bounding boxes is contained within a larger bounding box of the bounding boxes;
• expanding the smaller bounding box to generatean expanded bounding-box;
• and
• on condition that the expanded bounding box is still contained within the larger bounding box, merging the smaller bounding box into the larger bounding box to generate a merged bounding box;
• d) executing a second transformation to join at least some of the bounding boxes exhibiting sparse spacing to generate additional ones of the refined bounding boxes;
• e) applying an expansion transformation to the refined bounding boxes;
• and
• a classification network configured to utilize the refined bounding boxes to generate one or more classifications of one or more stamps depicted in a scanned image generated by the scanning element;
• and
• logic to apply the classifications to control the scanning element.

A copy protection method comprising:

• applying a digital image and a registered security stamp to a region proposal network configured to generate region proposals;
• refining the region proposals by:
• a) determining a first set of the region proposals exhibiting dense spacing;
• b) determining a second set of the region proposals exhibiting sparse spacing;
• c) executing a first transformation to merge at least some of the region proposals exhibiting dense spacing to generate refined region proposals;
• d) executing a second transformation to join at least some of the region proposals exhibiting sparse spacing to generate additional ones of the refined region proposals;
• e) applying an expansion transformation to the refined region proposals;
• and
• f) repeating a) through d) after applying the expansion transformation to the refined region proposals;
• operating a classification network on the refined region proposals to generate one or more classifications of one or more objects depicted in the region proposals;
• and
• applying the one or more classifications to control a document copy device.