Image processing apparatus, image processing system, image processing method, and storage medium
There is provided with an image processing apparatus, for example for image recognition such as object counting with machine learning. A generation unit, based on a first captured image, generates a first training data that indicates a first training image and an image recognition result for the first training image. A training unit, by performing training using the first training data, generates a discriminator for image recognition based on both the first training data and second training data that is prepared in advance and that indicates a second training image and an image recognition result for the second training image.
1. An image processing apparatus, comprising:
a generation unit configured to, based on a first captured image obtained under a first image capturing condition, generate a first training data that includes a first training image and an image recognition result for the first training image; and
a training unit configured to, by performing training using the first training data, generate a discriminator for image recognition on a captured image obtained under the first image capturing condition, based on both the first training data and second training data that is prepared in advance and that includes a second training image and an image recognition result for the second training image,
wherein the second training image is generated based on a second captured image that is obtained under a second image capturing condition that is different from the first image capturing condition.
2. The image processing apparatus according to claim 1 , wherein the training unit is further configured to perform training of the discriminator by using both the first training data and the second training data.
3. The image processing apparatus according to claim 1 , wherein the training unit, by using the first training data, performs incremental training of the discriminator which has been trained using the second training data.
4. The image processing apparatus according to claim 3 , wherein the training unit is further configured to perform the incremental training of the discriminator using mixed training data having mixture of the first training data and the second training data.
5. The image processing apparatus according to claim 4 , wherein
the first training data comprises a plurality of first data items each indicating a training image and a label thereof;
the second training data comprises a plurality of second data items each indicating a training image and a label thereof;
the mixed training data comprises a plurality of data items including at least one first data item and at least one second data item; and
a sum of the number of the first data items having a particular label in the mixed training data and the number of the second data items having the particular label in the mixed training data is the same as the number of the second data items having the particular label in the second training data.
6. The image processing apparatus according to claim 1 , wherein the discriminator is a discriminator for estimating a number of a counting target in an image, and the image recognition result of the first training image is the number of the counting target in the first training image, and the image recognition result of the second training image is the number of the counting target in the second training image.
7. The image processing apparatus according to claim 6 , wherein the generation unit is further configured to generate the first training image from a background image of the first captured image.
8. The image processing apparatus according to claim 6 , wherein the generation unit is further configured to extract the first training image from the first captured image, and to determine the number of the counting target in the first training image by using position information of the counting target in the first captured image.
9. The image processing apparatus according to claim 8 , wherein the generation unit is further configured to generate position information of the counting target in the first captured image by performing detection processing of the counting target on the first captured image.
10. The image processing apparatus according to claim 8 , wherein the generation unit is further configured to determine density information of the counting target for each region in the first captured image, and to extract the first training image from the first captured image in accordance with the density information for each region.
11. The image processing apparatus according to claim 1 , wherein
the generation unit is further configured to repeatedly generate the first training data, and the training unit is further configured to repeatedly perform training using the first training data, and
the image processing apparatus further comprises a control unit configured to determine whether or not to end repetition of the training based on at least one of a period at which training is repeated, a repetition count for the training, or a discrimination accuracy of the discriminator for which the training has been performed.
12. The image processing apparatus according to claim 1 , further comprising a recognition unit configured to perform image recognition using the discriminator generated by the training unit.
13. The image processing apparatus according to claim 1 , wherein the first or second image capturing conditions include at least one of an image capturing apparatus, an image capturing time period, an image capturing position, an image capturing angle of view, an image capturing distance in relation to a ground, or an image capturing angle in relation to a ground.
14. The image processing apparatus according to claim 1 , further comprising a status display unit configured to display a status of training by the training unit.
15. The image processing apparatus according to claim 14 , wherein the status display unit is further configured to display at least one of an elapsed time from start of training, the number of training data items used, an estimation error according to the discriminator, or training progress, as the status of training.
16. An image processing method, comprising:
based on a first captured image obtained under a first image capturing condition, generating a first training data that includes a first training image and an image recognition result for the first training image; and
by performing training using the first training data, generating a discriminator for image recognition on a captured image obtained under the first image capturing condition, based on both the first training data and second training data that is prepared in advance and that includes a second training image and an image recognition result for the second training image,
wherein the second training image is generated based on a second captured image that is obtained under a second image capturing condition that is different from the first image capturing condition.
17. A non-transitory storage medium storing a program which, when executed by a computer comprising a processor and a memory, causes the computer to:
based on a first captured image obtained under a first image capturing condition, generating a first training data that includes a first training image and an image recognition result for the first training image; and
by performing training using the first training data, generating a discriminator for image recognition on a captured image obtained under the first image capturing condition, based on both the first training data and second training data that is prepared in advance and that includes a second training image and an image recognition result for the second training image,
wherein the second training image is generated based on a second captured image that is obtained under a second image capturing condition that is different from the first image capturing condition.