Information processing apparatus, information processing method, and non-transitory computer-readable storage medium

An information processing apparatus comprising:

• a recognition processing unit configured to acquire a result of recognition processing of a convolutional neural network using training data including information of a training object, as an input;
• a region of interest setting unit configured to set a region of interest for the training data and/or an intermediate layer output of the convolutional neural network;
• a detail recognition unit configured to perform detail recognition processing more detailed than the recognition processing for the training data and/or the intermediate layer output in the region of interest;
• an integration processing unit configured to perform integration processing of a result of the detail recognition processing and the intermediate layer output;
• an intermediate input processing unit configured to input a result of the integration processing as the intermediate layer output to the convolutional neural network;
• and
• a learning unit configured to perform training processing concerning at least one of the recognition processing unit, the region of interest setting unit, the detail recognition unit, the integration processing unit, and the intermediate input processing unit.

An information processing method comprising:

• acquiring a result of recognition processing of a convolutional neural network using recognition object data including information of a recognition object, as an input;
• setting a region of interest for the recognition object data and/or an intermediate layer output of the convolutional neural network;
• performing detail recognition processing more detailed than the recognition processing for the recognition object data and/or the intermediate layer output in the region of interest;
• performing integration processing of a result of the detail recognition processing and the intermediate layer output;
• inputting a result of the integration processing as the intermediate layer output to the convolutional neural network;
• and
• outputting a result of the recognition processing.

An information processing method comprising:

• acquiring a result of recognition processing of a convolutional neural network using training data including information of a training object, as an input;
• setting a region of interest for the training data and/or an intermediate layer output of the convolutional neural network;
• performing detail recognition processing more detailed than the recognition processing for the training data and/or the intermediate layer output in the region of interest;
• performing integration processing of a result of the detail recognition processing and the intermediate layer output;
• inputting a result of the integration processing as the intermediate layer output to the convolutional neural network;
• and
• performing training processing concerning at least one of the recognition processing, the setting of the region of interest, the detail recognition processing, the integration processing, and the input.

A non-transitory computer-readable storage medium storing a computer program for causing a computer to function as:

• a recognition processing unit configured to acquire a result of recognition processing of a convolutional neural network using recognition object data including information of a recognition object, as an input;
• a region of interest setting unit configured to set a region of interest for the recognition object data and/or an intermediate layer output of the convolutional neural network;
• a detail recognition unit configured to perform detail recognition processing more detailed than the recognition processing for the recognition object data and/or the intermediate layer output in the region of interest;
• an integration processing unit configured to perform integration processing of a result of the detail recognition processing and the intermediate layer output;
• an intermediate input processing unit configured to input a result of the integration processing as the intermediate layer output to the convolutional neural network;
• and
• an output unit configured to output a result of the recognition processing.

A non-transitory computer-readable storage medium storing a computer program for causing a computer to function as:

• a recognition processing unit configured to acquire a result of recognition processing of a convolutional neural network using training data including information of a training object, as an input;
• a region of interest setting unit configured to set a region of interest for the training data and/or an intermediate layer output of the convolutional neural network;
• a detail recognition unit configured to perform detail recognition processing more detailed than the recognition processing for the training data and/or the intermediate layer output in the region of interest;
• an integration processing unit configured to perform integration processing of a result of the detail recognition processing and the intermediate layer output;
• an intermediate input processing unit configured to input a result of the integration processing as the intermediate layer output to the convolutional neural network;
• and
• a learning unit configured to perform training processing concerning at least one of the recognition processing unit, the region of interest setting unit, the detail recognition unit, the integration processing unit, and the intermediate input processing unit.