Abstract: An image processing apparatus applies a learning process to a learning model using training data generated from first original images obtained by capturing a subject of inspection. The apparatus then performs, using the learned learning model, deduction processing on input data generated from a second original image obtained by capturing a subject of inspection. The training data is generated using a plurality of partial images from each of the first original images, and the deduction process uses a partial image of the second original image as the input data. The detection rate of an abnormal portion having a small size included in images of the subject of inspection is thereby improved.