Image determination device, training method and non-transitory computer readable medium storing program
Provided is an image determination device. The image determination device is provided with: feature extractors which output, on the basis of an image to be examined, each piece of feature data indicating a specific feature of the image; a determiner which outputs, on the basis of the feature data output from the extractors, output data indicating the determination result pertaining to the image; and a training part which trains the determiner so as to output, output data indicating the label data associated with the training image on the basis of the feature data output when the training image is input to the extractors, wherein the training part further trains, by using new training data, the determiner so that the output data indicating the label data associated with the image is output by the determiner.
1. An image determination device comprising:
one or a plurality of feature extractors each of which outputs feature data indicating a specific feature of an image to be examined on the basis of the image;
a determiner which outputs output data indicating a determination result related to the image on the basis of the feature data output from the one or plurality of feature extractors; and
a training part which causes the determiner to train so as to output the output data indicating label data associated with a training image, on the basis of the feature data which is output in a case where the training image is input to the one or plurality of feature extractors by using training data including the training image and the label data,
wherein the training part performs additional training of the determiner without making a change to the one or plurality of feature extractors to determine which the one or plurality of feature extractors are focused on to obtain the output data so that the output data indicating the label data associated with the image by the determiner is output, on the basis of the feature data which is output in a case where the image to be examined included in new training data is input to the one or plurality of feature extractors by using the new training data.
2. The image determination device according to claim 1 , wherein the training part performs the additional training of the determiner by executing at least any one of updating of parameters of the determiner and updating of a weight of the feature data which is input to the determiner.
3. The image determination device according to claim 2 , wherein the training part updates parameters of some feature extractors of the one or plurality of feature extractors together with the additional training of the determiner.
4. The image determination device according to claim 2 , further comprising:
an editing part which edits the one or plurality of feature extractors on the basis of an input.
5. The image determination device according to claim 1 , wherein the training part updates parameters of some feature extractors of the one or plurality of feature extractors together with the additional training of the determiner.
6. The image determination device according to claim 5 , further comprising:
an editing part which edits the one or plurality of feature extractors on the basis of an input.
7. The image determination device according to claim 1 , further comprising:
an editing part which edits the one or plurality of feature extractors on the basis of an input.
8. The image determination device according to claim 7 , wherein the editing part adds a new feature extractor that outputs the feature data indicating the specific feature of the image on the basis of the input.
9. The image determination device according to claim 8 , wherein the editing part deletes some of the one or plurality of feature extractors on the basis of the input.
10. The image determination device according to claim 8 , wherein the editing part edits parameters of the one or plurality of feature extractors on the basis of the input.
11. The image determination device according to claim 7 , wherein the editing part deletes some of the one or plurality of feature extractors on the basis of the input.
12. The image determination device according to claim 11 , wherein the editing part edits parameters of the one or plurality of feature extractors on the basis of the input.
13. The image determination device according to claim 7 , wherein the editing part edits parameters of the one or plurality of feature extractors on the basis of the input.
14. A training method comprising:
causing a determiner to train so as to output output data indicating label data associated with a training image, on the basis of feature data which is output in a case where the training image is input to one or a plurality of feature extractors by using training data including the training image and the label data, wherein the determiner outputting the output data indicating a determination result related to an image to be examined, on the basis of the feature data output from the one or plurality of feature extractors each outputting the feature data indicating a specific feature of the image on the basis of the image; and
causing the training part to perform additional training of the determiner without making a change to the one or plurality of feature extractors to determine which the one or plurality of feature extractors are focused on to obtain the output data so that the output data indicating the label data associated with the image by the determiner is output, on the basis of the feature data which is output in a case where the image to be examined included in new training data is input to the one or plurality of feature extractors by using the new training data.
15. A non-transitory computer readable medium storing an image determination program causing a computation part included in an image determination device to function as:
one or a plurality of feature extractors each of which outputs feature data indicating a specific feature of an image to be examined on the basis of the image;
a determiner which outputs output data indicating a determination result related to the image on the basis of the feature data output from the one or plurality of feature extractors; and
a training part which causes the determiner to train so as to output the output data indicating label data associated with a training image, on the basis of the feature data which is output in a case where the training image is input to the one or plurality of feature extractors by using training data including the training image and the label data,
wherein the training part performs additional training of the determiner without making a change to the one or plurality of feature extractors to determine which the one or plurality of feature extractors are focused on to obtain the output data so that the output data indicating the label data associated with the image by the determiner is output, on the basis of the feature data which is output in a case where the image to be examined included in new training data is input to the one or plurality of feature extractors by using the new training data.