Learning device, learning method and storage medium
A learning device is configured to acquire first training data that is training data regarding a registered category object and second training data regarding an unregistered category object, the registered category object belonging to a registered category which is registered as an identification target, the unregistered category object not belonging to the registered category. The learning device is also configured to calculate, on a basis of the first training data and the second training data, an identification score indicating a degree of certainty that an object to be identified belongs to the registered category. The learning device is also configured to calculate, on a basis of the second training data, an unregistered score indicating a degree of certainty that the object to be identified does not belong to the registered category. The learning device is also configured to learn an identification model which performs an identification regarding the registered category based on the identification score and the unregistered score.
1. A learning device comprising a processor configured to:
acquire first training data that is training data regarding a registered category object and second training data regarding an unregistered category object, the registered category object belonging to a registered category which is registered as an identification target, the unregistered category object not belonging to the registered category;
calculate, on a basis of the first training data and the second training data, an identification score indicating a degree of certainty that an object to be identified belongs to the registered category;
calculate, on a basis of the second training data, an unregistered score indicating a degree of certainty that the object to be identified does not belong to the registered category; and
learn an identification model which performs an identification regarding the registered category based on the identification score and the unregistered score.
2. The learning device according to claim 1 ,
wherein the processor is configured to calculate the unregistered score based on the identification score calculated based on the second training data.
3. The learning device according to claim 2 ,
wherein the processor is configured to calculate the unregistered score based on the identification score that is the maximum among the identification score calculated for each registration category.
4. The learning device according to claim 2 ,
wherein the processor is configured to calculate the unregistered score based on
a sum of the identification score calculated for each registered category or
a weighted sum of the identification score calculated for each registered category weighted based on a ranking of a magnitude of the identification score.
5. The learning device according to claim 2 ,
wherein the processor is configured to calculate the unregistered score based on a product of the identification score calculated for each registered category.
6. The learning device according to claim 1 ,
wherein the processor is configured to determine the unregistered score to be a value obtained by decreasing a candidate value of the unregistered score based on a constant.
7. The learning device according to claim 1
wherein the processor is configured, in learning the identification model to:
calculate
a first loss that is a loss for the identification score calculated from the first training data and
a second loss that is a loss for the unregistered score calculated from the second training data; and
determine a parameter of the identification model based on the first loss and the second loss.
8. The learning device according to claim 1 ,
wherein the processor is configured to perform the learning of the identification model which is used for calculating the identification score.
9. A learning method comprising:
acquiring first training data that is training data regarding a registered category object and second training data regarding an unregistered category object, the registered category object belonging to a registered category which is registered as an identification target, the unregistered category object not belonging to the registered category;
calculating, on a basis of the first training data and the second training data, an identification score indicating a degree of certainty that an object to be identified belongs to the registered category;
calculating, on a basis of the second training data, an unregistered score indicating a degree of certainty that the object to be identified does not belong to the registered category; and
learning an identification model which performs an identification regarding the registered category based on the identification score and the unregistered score.
10. A non-transitory computer readable storage medium storing a program executed by a computer, the program causing the computer to:
acquire first training data that is training data regarding a registered category object and second training data regarding an unregistered category object, the registered category object belonging to a registered category which is registered as an identification target, the unregistered category object not belonging to the registered category;
calculate, on a basis of the first training data and the second training data, an identification score indicating a degree of certainty that an object to be identified belongs to the registered category;
calculate, on a basis of the second training data, an unregistered score indicating a degree of certainty that the object to be identified does not belong to the registered category; and
learn an identification model which performs an identification regarding the registered category based on the identification score and the unregistered score.