Learning device, learning method, identifying device, identifying method, program, and information processing system

A learning method of a learning device for learning an identifier for identifying a predetermined object to be identified, with said learning device including a processor, said learning method comprising:

• extracting, with said processor, a feature point from each of a plurality of images for generation to be used for generating an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, which are made up of a positive image where said object to be identified exists, and a negative image where said object to be identified does not exist;
• extracting, with said processor feature point feature amount representing the feature of said feature point from said image for generation;
• calculating, with said processor the whole feature amount representing the feature of the whole of said image for generation based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation, of the whole range on said image for generation;
• and
• generating, with said processor, said identifier based on said whole feature amount of said image for generation, and a correct answer label representing whether said image for generation is said positive image or said negative image.

A non-transitory computer-readable medium including a program, which when executed by a computer, causes the computer to perform a method, the method comprising:

• extracting a feature point from each of a plurality of images for generation to be used for generating an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, which are made up of a positive image where said object to be identified exists, and a negative image where said object to be identified does not exist;
• extracting a feature point feature amount representing the feature of said feature point from said image for generation;
• calculating a whole feature amount representing the feature of the whole of said image for generation based on the feature point feature amount of a feature point existing on a feature point selection range to determined based on said plurality of images for generation, of the whole range on said image for generation;
• and
• generating said identifier based on said whole feature amount of said image for generation, and a correct answer label representing whether said image for generation is said positive image or said negative image.

An identifying device comprising:

• feature point extracting means configured to extract a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified;
• feature point feature amount extracting means configured to extract feature point feature amount representing the feature of said feature point from said image to be processed;
• whole feature amount calculating means configured to calculate the whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on a plurality of images for generation to be used for learning for generating an identifier for identifying each different object to be identified of the whole range on said image to be processed;
• and
• identifying means configured to identify, based on an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, and said whole feature amount, whether or not a subject existing on said image to be processed is a predetermined object to be identified.

An identifying method of an identifying device for identifying whether or not a subject existing on an image is a predetermined object to be identified, with said identifying device including a processor, said identifying method comprising:

• extracting, with said processor, a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified;
• extracting, with said processor, a feature point feature amount representing the feature of said feature point from said image to be processed;
• calculating, with said processor, the whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation to be used for learning for generating an identifier which identifies each different object to be identified, of the whole range on said image to be processed;
• and
• identifying, with said processor, whether or not a subject existing on said image to be processed is a predetermined object to be identified, based on an identifier for identifying whether or not a subject exiting on an image is a predetermined object to be identified, and said whole feature amount.

A non-transitory computer-readable medium including a program, which when executed by a computer, causes the computer to perform a method, the method comprising:

• extracting a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified;
• extracting a feature point feature amount representing the feature of said feature point from said image to be processed;
• calculating a whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation to be used for learning for generating an identifier which identifies each different object to be identified, of the whole range on said image to be processed;
• and
• identifying whether or not a subject existing on said image to be processed is a predetermined object to be identified, based on an identifier for identifying whether or not a subject exiting on an image is a predetermined object to be identified, and said whole feature amount.

An information processing system comprising:

• a learning device configured to learn an identifier for identifying a predetermined object to be identified;
• and
• an identifying device configured to identify whether or not a subject existing on an image is a predetermined object to be identified, using said identifier;
• wherein said learning device includes
• first feature point extracting means configured to extract a feature point from each of a plurality of images for generation to be used for generating an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, which are made up of a positive image where said object to be identified exists, and a negative image where said object to be identified does not exist,
• first feature point feature amount extracting means configured to extract feature point feature amount representing the feature of said feature point from said image for generation,
• first whole feature amount calculating means configured to calculate the whole feature amount representing the feature of the whole of said image for generation based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation, of the whole range on said image for generation, and
• identifier generating means configured to generate said identifier based on said whole feature amount of said image for generation, and a correct answer label representing whether said image for generation is said positive image or said negative image;
• and wherein said identifying device includes
• second feature point extracting means configured to extract a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified,
• second feature point feature amount extracting means configured to extract feature point feature amount representing the feature of said feature point from said image to be processed,
• second whole feature amount calculating means configured to calculate the whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on a plurality of images for generation to be used for learning by said learning device, of the whole range on said image to be processed, and
• identifying means configured to identify, based on said identifier generated by said identifier generating means, and said whole feature amount of said image to be processed, whether or not a subject existing on said image to be processed is a predetermined object to be identified.

A learning device comprising:

• a feature point extracting unit configured to extract a feature point from each of a plurality of images for generation to be used for generating an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, which are made up of a positive image where said object to be identified exists, and a negative image where said object to be identified does not exist;
• a feature point feature amount extracting unit configured to extract feature point feature amount representing the feature of said feature point from said image for generation;
• a whole feature amount calculating unit configured to calculate the whole feature amount representing the feature of the whole of said image for generation based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation, of the whole range on said image for generation;
• and
• an identifier generating unit configured to generate said identifier based on said whole feature amount of said image for generation, and a correct answer label representing whether said image for generation is said positive image or said negative image.

An identifying device comprising:

• a feature point extracting unit configured to extract a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified;
• a feature point feature amount extracting unit configured to extract feature point feature amount representing the feature of said feature point from said image to be processed;
• a whole feature amount calculating unit configured to calculate the whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on a plurality of images for generation to be used for learning for generating an identifier for identifying each different object to be identified of the whole range on said image to be processed;
• and
• an identifying unit configured to identify, based on an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, and said whole feature amount, whether or not a subject existing on said image to be processed is a predetermined object to be identified.

An information processing system comprising:

• a learning device configured to learn an identifier for identifying a predetermined object to be identified;
• and
• an identifying device configured to identify whether or not a subject existing on an image is a predetermined object to be identified, using said identifier;
• wherein said learning device includes
• a first feature point extracting unit configured to extract a feature point from each of a plurality of images for generation to be used for generating an identifier for identifying whether or not a subject existing on an image is a predetermined object to be identified, which are made up of a positive image where said object to be identified exists, and a negative image where said object to be identified does not exist,
• a first feature point feature amount extracting unit configured to extract feature point feature amount representing the feature of said feature point from said image for generation,
• a first whole feature amount calculating unit configured to calculate the whole feature amount representing the feature of the whole of said image for generation based on the feature point feature amount of a feature point existing on a feature point selection range determined based on said plurality of images for generation, of the whole range on said image for generation, and
• an identifier generating unit configured to generate said identifier based on said whole feature amount of said image for generation, and a correct answer label representing whether said image for generation is said positive image or said negative image;
• and wherein said identifying device includes
• a second feature point extracting unit configured to extract a feature point from an image to be processed serving as a processing object for identifying whether or not a subject existing on an image is a predetermined object to be identified,
• a second feature point feature amount extracting unit configured to extract feature point feature amount representing the feature of said feature point from said image to be processed,
• a second whole feature amount calculating unit configured to calculate the whole feature amount representing the feature of the whole of said image to be processed based on the feature point feature amount of a feature point existing on a feature point selection range determined based on a plurality of images for generation to be used for learning by said learning device, of the whole range on said image to be processed, and
• an identifying unit configured to identify, based on said identifier generated by said identifier generating unit, and said whole feature amount of said image to be processed, whether or not a subject existing on said image to be processed is a predetermined object to be identified.