Information processing apparatus, control method, and storage medium

A method for controlling an information processing apparatus comprising:

• receiving a plurality of input document images as learning data;
• deciding a reference image from the plurality of input document images,
• deciding a reference direction of the reference image;
• generating a rotated image by rotating the reference image to a predetermined direction different from the reference direction of the reference image;
• obtaining an image feature amount of the reference image and an image feature amount of the rotated image;
• generating, based on the obtained image feature amount of the reference image and the image feature amount of the rotated image, at least one direction determiner that determines whether a direction of an input document image is the reference direction of the reference image or a direction of the rotated image;
• automatically determining the direction of each of the plurality of input document images by inputting image feature amount of each of the input document images to the generated at least one direction determiner, wherein the direction determiner does not use an OCR processing;
• unifying the direction of each of the plurality of input document images with the reference direction of the reference image in accordance with the determined direction;
• and
• generating, based on image feature amount of the plurality of input document images whose directions have been unified, a classification rule to be used in classification of new document images that are different from the plurality of input document images received as the learning data.

A non-transitory storage medium on which is stored a computer program for making a computer execute a method for controlling an information processing apparatus, the method comprising:

• receiving a plurality of input document images as learning data;
• deciding a reference image from the plurality of input document images;
• deciding a reference direction of the reference image;
• generating a rotated image by rotating the reference image to a predetermined direction different from the reference direction of the reference image;
• obtaining an image feature amount of the reference image and an image feature amount of the rotated image;
• generating, based on the obtained image feature amount of the reference image and the image feature amount of the rotated image, at least one direction determiner that determines whether a direction of an input document image is the reference direction of the reference image or a direction of the rotated image;
• automatically determining the direction of each of the plurality of input document images by inputting image feature amount of each of the input document images to the generated at least one direction determiner, wherein the direction determiner does not use an OCR processing;
• unifying the direction of each of the plurality of input document images with the reference direction of the reference image in accordance with the determined direction;
• and
• generating, based on image feature amount of the plurality of input document images whose directions have been unified, a classification rule to be used in classification of new document images that are different from the plurality of input document images received as the learning data.