Image suggestion apparatus, image suggestion method, and image suggestion program

An image suggestion method comprising:

• collecting and storing data, which is collected for each user and is related to a combination of a first room and a first image, in a data storage unit, the data being related to the combination preferred by the user;
• generating information which indicates a relationship between the first room and the first image and in which a preference of the user is reflected by using the data stored in the data storage unit and corresponding to the user;
• acquiring a second image which is a second room image captured by the user;
• selecting a third image which is an image in which the preference of the user is reflected, from a registered image group or an image group owned by the user, based on the acquired second image and the generated information which indicates the relationship;
• and
• suggesting, as an image corresponding to an image print for decorating the second room of the user, the selected third image to the user,
• wherein the generating information is performed by a generation unit that includes a convolution neural network, a loss value calculation unit, and a parameter control unit,
• the convolution neural network includes an input layer, an intermediate layer, and an output layer,
• a fourth image, which is an image of the first room decorated with the first image, is input to the input layer,
• the intermediate layer includes a plurality of sets each including a convolution layer and a pooling layer, acquires a feature map, and outputs the acquired feature map to the output layer,
• the loss value calculation unit acquires the feature map that is output from the output layer of the convolution neural network and the first image, and calculates a loss value between the feature map and the first image,
• the parameter control unit adjusts the weight parameter of the convolution neural network based on the loss value calculated by the loss value calculation unit, and
• the generation unit acquires the weight parameter of the convolution neural network as the room-and-image relationship information.

A non-transitory computer readable recording medium storing an image suggestion program causing a computer to realize:

• a function of collecting and storing data, which is collected for each user and is related to a combination of a first room and a first image, in a data storage unit, the data being related to the combination preferred by the user;
• a function of generating information which indicates a relationship between the first room and the first image and in which a preference of the user is reflected by using the data stored in the data storage unit and corresponding to the user;
• a function of acquiring a second image which is a second room image captured by the user;
• a function of selecting a third image which is an image in which the preference of the user is reflected, from a registered image group or an image group owned by the user, based on the acquired second image and the generated information which indicates the relationship;
• and
• a function of suggesting, as an image corresponding to an image print for decorating the second room of the user, the selected third image to the user,
• wherein the function of generating information is performed by a generation unit that includes a convolution neural network, a loss value calculation unit, and a parameter control unit,
• the convolution neural network includes an input layer, an intermediate layer, and an output layer,
• a fourth image, which is an image of the first room decorated with the first image, is input to the input layer,
• the intermediate layer includes a plurality of sets each including a convolution layer and a pooling layer, acquires a feature map, and outputs the acquired feature map to the output layer,
• the loss value calculation unit acquires the feature map that is output from the output layer of the convolution neural network and the first image, and calculates a loss value between the feature map and the first image,
• the parameter control unit adjusts the weight parameter of the convolution neural network based on the loss value calculated by the loss value calculation unit, and
• the generation unit acquires the weight parameter of the convolution neural network as the room-and-image relationship information.