Computer-readable recording medium having stored therein information processing program, method for processing information, and information processing apparatus

A method for processing information executed by a computer, the method comprising:

• acquiring a training data set including a plurality of pieces of training data, each of the plurality pieces of training data including an image obtained by photographing a training target, first annotation data representing a rectangular region including the training target in the image, and second annotation data representing one or more pixels corresponding to the training target in the image;
• training, based on the image and the first annotation data included in the training data set, an object detection model included in an estimation model, the object detection model specifying a rectangular region including the training target in the image in response to inputting of the image;
• training, based on the image and the second annotation data included in the training data set, a neural network included in the estimation model, the neural network outputs one or more pixels estimated to correspond to the training target in the image in response to inputting of the image;
• and
• calculating a first index value related to a relationship of a pixel number between a pixel number of the rectangular region including the training target specified by the first annotation data and a pixel number corresponding to the training target specified by the second annotation data, the estimation model after the training and the calculated first index value being used in a determination process that determines, based on the calculated first index value and a second index value, whether or not a target that is to be determined and that is included in a target image is normal, the second index value being related to a relationship between a pixel number corresponding to a rectangular region specified by an estimation result obtained by using the trained object detection model in an output result and a pixel number corresponding to the target specified by an estimation result obtained by using the trained neural network, the output result being output from the estimation model after the training in response to inputting of the target image including the target.

An information processing apparatus comprising:

• a memory;
• and
• a processor coupled to the memory, the processor being configured to:
• acquire a training data set including a plurality of pieces of training data, each of the plurality pieces of training data including an image obtained by photographing a training target, first annotation data representing a rectangular region including the training target in the image, and second annotation data representing one or more pixels corresponding to the training target in the image;
• train based on the image and the first annotation data included in the training data set, an object detection model included in an estimation model, the object detection model specifying a rectangular region including the training target in the image in response to inputting of the image;
• train, based on the image and the second annotation data included in the training data set, a neural network included in the estimation model, the neural network outputs one or more pixels estimated to correspond to the training target in the image in response to inputting of the image;
• calculate a first index value related to a relationship of a pixel number between a pixel number of the rectangular region including the training target specified by the first annotation data and a pixel number corresponding to the training target specified by the second annotation data;
• and
• execute a determination process that determines, based on the calculated first index value and a second index value, whether or not a target that is to be determined and that is included in a target image is normal, the second index value being related to a relationship between a pixel number corresponding to a rectangular region specified by an estimation result obtained by using the trained object detection model in an output result and a pixel number corresponding to the target specified by an estimation result obtained by using the trained neural network, using the estimation model after the training and the calculated first index value, the output result being output from the estimation model after the training in response to inputting of the target image including the target.