Training device, training method and non-transitory computer readable medium storing training program
A training device includes a data acquirer that acquires training data including a first image of a first type and a second image of a second type that is different from the first type, a first trainer that repeats an epoch for causing the first image and the second image to be learned under a same condition multiple times, and a second trainer that repeats an epoch for training a trained learning model generated by the first trainer with a rate of error back-propagation in regard to the first image and a rate of error back-propagation in regard to the second image being made different from each other.
1 . A training device comprising:
a data acquirer that acquires training data including a first image of a first type and a second image of a second type that is different from the first type;
a first trainer that repeats an epoch for causing the first image and the second image to be learned under a same condition multiple times; and
a second trainer that repeats an epoch for training a trained learning model generated by the first trainer with a rate of error back-propagation in regard to the first image and a rate of error back-propagation in regard to the second image being made different from each other.
2 . The training device according to claim 1 , wherein
the first trainer causes the first image and the second image to be learned with a ratio of error back-propagation in regard to the first image and a ratio of error back-propagation in regard to the second image being same.
3 . The training device according to claim 1 , wherein
the training data includes a plurality of sets of an inspection image obtained when an image of an inspection subject is picked up as the first image, a first inspection result obtained when each inspection image is inspected by an inspection device and a second inspection result obtained when the inspection subject is inspected with a method different from the inspection device,
the first image is the inspection image the first inspection result of which represents presence of a defect and the second inspection result of which represents presence of a defect, and
the second image is the inspection image the first inspection result of which represents presence of a defect and the second inspection result of which represents absence of a defect.
4 . The training device according to claim 2 , wherein
the second trainer sets a value smaller than a rate of error back-propagation in regard to the first image as a rate of error back-propagation in regard to the second image.
5 . The training device according to claim 1 , further comprising a display that displays a deviation representing a difference between an evaluation result obtained when the training data is input during training to a learning model that is caused to learn by the first trainer and an evaluation result obtained when a plurality of test images obtained when images of a plurality of inspection subjects are respectively picked up are input during training to a learning model that is caused to learn by the first trainer.
6 . The training device according to claim 1 , further comprising a count determiner that determines a training repetition count of the first trainer based on an evaluation result obtained when a plurality of test images are input during training to a learning model that is caused to learn by the first trainer, the plurality of test images being obtained when images of a plurality of inspection subjects are picked up.
7 . The training device according to claim 6 , wherein
the first trainer compares a change amount of an evaluation result in regard to the repetition count with a predetermined value.
8 . A training method including:
a data acquiring step of acquiring training data including a first image of a first type and a second image of a second type that is different from the first type;
a first training step of repeating an epoch for causing the first image and the second image to be learned under a same condition multiple times; and
a second training step of repeating an epoch for training a trained learning model generated in the first training step with a rate of error back-propagation in regard to the first image and a rate of error back-propagation in regard to the second image being made different from each other.
9 . A non-transitory computer readable medium storing a training program that causes a computer to execute:
a data acquiring step of acquiring training data including a first image of a first type and a second image of a second type that is different from the first type;
a first training step of repeating an epoch for causing the first image and the second image to be learned under a same condition multiple times; and
a second training step of repeating an epoch for training a trained learning model generated in the first training step with a rate of error back-propagation in regard to the first image and a rate of error back-propagation in regard to the second image being made different from each other.