Learning model generating device, and type identification system for generating learning model and using generated learning model to infer type of image defect

A learning model generating device capable of learning a data set containing image defect-containing training prints prepared for each type of image defect to generate a learning model, the learning model generating device comprising:

• a first image reading device that reads images of the training prints;
• and
• a first control device that includes a processor and functions, through the processor executing a first control program, as:
• a first segmenter that segments, on a predetermined first segmentation condition, each of the images of the training prints obtained by reading performed by the first image reading device;
• a learning model generator that learns segmented images obtained by segmentation performed by the first segmenter to generate a first learning model for use in inferring a type of an image defect;
• and
• a first compressor that compresses, on a predetermined first compression condition, each of the images of the training prints obtained by reading performed by the first image reading device,
• wherein the first segmenter segments, on a predetermined second segmentation condition, each of compressed images obtained by compression performed by the first compressor, and
• the learning model generator learns compressed and segmented images obtained by compression performed by the first compressor and segmentation performed by the first segmenter to generate a second learning model for use in inferring a type of an image defect,
• wherein the first compressor further compresses, on a predetermined third compression condition, each of the compressed images of the training prints obtained by previous compression,
• the first segmenter segments, on a predetermined third segmentation condition, each of compressed images of the training prints obtained by further compression performed by the first compressor, and
• the learning model generator learns compressed and segmented images obtained by the further compression performed by the first compressor and segmentation performed by the first segmenter to generate a third learning model for use in inferring a type of an image defect.

A type identification system comprising a learning model generating device capable of learning a data set containing image defect-containing training prints prepared for each type of image defect to generate a learning model and a type identification device using the first learning model and the second learning model both generated by the learning model generating device to identify a type of an image defect contained on a print to be inspected,

• the learning model generating device comprising:
• a first image reading device that reads images of the training prints;
• and
• a first control device that includes a processor and functions, through the processor executing a first control program, as:
• a first segmenter that segments, on a predetermined first segmentation condition, each of the images of the training prints obtained by reading performed by the first image reading device;
• a learning model generator that learns segmented images obtained by segmentation performed by the first segmenter to generate a first learning model for use in inferring a type of an image defect;
• and
• a first compressor that compresses, on a predetermined first compression condition, each of the images of the training prints obtained by reading performed by the first image reading device,
• wherein the first segmenter segments, on a predetermined second segmentation condition, each of compressed images obtained by compression performed by the first compressor, and
• the learning model generator learns compressed and segmented images obtained by compression performed by the first compressor and segmentation performed by the first segmenter to generate a second learning model for use in inferring a type of an image defect,
• the type identification device comprising:
• a second image reading device that reads an image of the print to be inspected;
• and
• a second control device that includes a processor and functions, through the processor executing a second control program, as:
• a second segmenter that segments, on the same segmentation condition as the first segmentation condition, the image of the print to be inspected obtained by reading performed by the second image reading device;
• an inferrer that uses the first learning model to infer a type of an image defect to which each of segmented images obtained by segmentation performed by the second segmenter applies;
• and
• a second compressor that compresses, on the same compression condition as the first compression condition, the image of the print to be inspected,
• wherein the second segmenter segments, on the same segmentation condition as the second segmentation condition, a compressed image obtained by compression performed by the second compressor, and
• the inferrer uses the second learning model to infer a type of an image defect to which each of compressed and segmented images obtained by compression performed by the second compressor and segmentation performed by the second segmenter applies.