A method for building an image determination model that determines whether a print image is a good image or a failure image includes a step of determining whether the print image is a high-density image or a low-density image; a step of generating teaching data by combining pseudo-failure data with to-be-combined image data, based on a result of the determination; and a step of performing machine learning. Regarding the combining of the pseudo-failure data, for a print image determined to be the high-density image, low-density pseudo-failure data is combined with to-be-combined image data, and for a print image determined to be the low-density image, high-density pseudo-failure data is combined with to-be-combined image data.