Training device and training method for training image processing device
The disclosure relates to a training device and method for an image processing device and an image processing device. The training device is used for training first and second image processing units, comprising: a training unit to input a first realistic image without a specific feature into the first image processing unit to generate a first generated image with the specific feature through first image processing, and to input a second realistic image with the specific feature into the second image processing unit to generate a second generated image without the specific feature through second image processing; and a classifying unit performing classification processing to discriminate realistic and generated images, wherein the training unit performs first training processing of training the classifying unit based on the realistic and generated images, and performs second training processing of training the first and second image processing units based on the training result.
1. A training device to train a first image process and a second image process, the training device comprising:
a memory; and
a processor coupled to the memory and the processor is to,
perform a training process to input a first realistic image without a specific feature into the first image process to generate a first generated image with the specific feature through first image processing, and to input a second realistic image with the specific feature into the second image process to generate a second generated image without the specific feature through second image processing; and
perform a classification process to discriminate realistic images including the first realistic image and the second realistic image and generated images including the first generated image and the second generated image,
wherein the training process is to,
perform a first training process of training the classification process based on the realistic images and the generated images; and
perform a second training process of training the first image process and the second image process based on a training result of the first training processing to reduce classification accuracy of the classification process, and
wherein,
the first image process is to generate a first residual image, the first generated image is a sum of the first realistic image and the first residual image, and
the second image process is to generate a second residual image, the second generated image is a sum of the second realistic image and the second residual image.
2. The training device according to claim 1 , wherein
the training process is to alternately perform the first training process and the second training process until the classification accuracy of the classification process is stable.
3. The training device according to claim 1 , wherein
the first residual image and the second residual image are related only to the specific feature.
4. The training device according to claim 1 , wherein
the classification process is to discriminate the first realistic image and the second realistic image.
5. The training device according to claim 1 , wherein
the first image process is to generate a third generated image based on the second generated image,
the second image process is to generate a fourth generated image based on the first generated image,
the training process is to perform a third training process of training the first image process and the second image process using the first realistic image, the second realistic image, the third generated image and the fourth generated image.
6. The training device according to claim 5 , wherein
the classification process is to alternately perform the first training process, the second training processing and the third training process until the classification accuracy of the classification process is stable.
7. The training device according to claim 5 , wherein
the third training process trains the first image process and the second image process using a difference between the first realistic image and the fourth generated image and a difference between the second realistic image and the third generated image.
8. An image processing device comprising a memory, and a processor coupled to the memory and the processor is to implement the first image process and the second image process, according to claim 1 .
9. A training method by a training device including a memory and a processor coupled to the memory, to train a first image process and a second image process of an image processing device, the training method comprising:
by the training device,
inputting a first realistic image without a specific feature into the first image process to generate a first residual image and to generate a first generated image with the specific feature through first image processing according to a sum of the first realistic image and the first residual image;
inputting a second realistic image with the specific feature into the second image process to generate a second residual image and to generate a second generated image without the specific feature through second image processing according to a sum of the second realistic image and the second residual image;
performing a classification processing to discriminate realistic images including the first realistic image and the second realistic image and generated images including the first generated image and the second generated image, and
performing a first training process to train the classification process, based on the realistic images and the generated images, and performing a second training process to train the first image process and the second image process based on a training result of the first training process to reduce classification accuracy of the classification process.