Machine learning devices for images, image processing systems equipped with machine learning devices for images, and machine learning methods for images.

Abstract

The machine learning apparatus of the present invention comprises: a dataset memory unit storing a first dataset; a first model learned using the first dataset; a degradation detection unit detecting degradation of the inference result inferred by the first model for an inference object image from an external device; a dataset processing unit processing the first dataset to obtain a second dataset based on the degradation detection result of the degradation detection unit; a relearning dataset memory unit storing the second dataset; and a second model learned using the second dataset; and further comprises: a differential verification unit acquiring a first inference result and a second inference result obtained by inferring the data in the dataset memory unit and the relearning dataset memory unit using the first model, and a third inference result and a fourth inference result obtained by inferring the data in the dataset memory unit and the relearning dataset memory unit using the second model, and providing evaluation results of the changes between the first inference result and the second inference result, and the changes between the third inference result and the fourth inference result.

Images