Generation method and information processing apparatus
An information processing apparatus calculates a parameter for each of a plurality of layers included in a first neural network through machine learning using a plurality of image datasets each containing a human biometric image. The information processing apparatus generates a determination model of determining the authenticity of a human biometric image included in a received image dataset, by setting a parameter calculated for a first layer of the first neural network in the first layer included in a second neural network that includes the first layer and does not include a second layer.
1 . A generation method comprising:
calculating, by a processor, a parameter for each of a plurality of layers included in a first neural network through first machine learning using a plurality of image datasets each containing a human biometric image, the plurality of layers including a first layer and a second layer; and
generating, by the processor, a determination model of determining authenticity of a human biometric image included in a received image dataset, by setting a parameter calculated for the first layer of the first neural network in the first layer included in a second neural network that includes the first layer and does not include the second layer.
2 . The generation method according to claim 1 , wherein
the generating of the determination model includes performing second machine learning of updating the second neural network using the set parameter as an initial value, and
the determination model is the updated second neural network.
3 . The generation method according to claim 1 , wherein the second layer is placed after the first layer.
4 . The generation method according to claim 1 , wherein
the plurality of layers include
a plurality of first layers including the first layer,
a plurality of second layers including the second layer and being placed after the plurality of first layers, and
a plurality of third layers placed before the plurality of first layers, and
the second neural network includes the plurality of first layers and does not include the plurality of second layers among the plurality of layers.
5 . The generation method according to claim 4 , wherein the second neural network includes at least one third layer among the plurality of third layers and does not include remaining third layers other than the at least one third layer.
6 . The generation method according to claim 1 , wherein
the plurality of image datasets used in the first machine learning are generated by dividing first video data containing a plurality of frames into sets of a predetermined number of frames each, and
the received image dataset that the determination model receives is generated by extracting a predetermined number of frames from beginning of second video data.
7 . An information processing apparatus comprising:
a memory that stores a plurality of image datasets each containing a human biometric image; and
a processor coupled to the memory and the processor configured to
calculate a parameter for each of a plurality of layers included in a first neural network through first machine learning using the plurality of image datasets, the plurality of layers including a first layer and a second layer, and
generate a determination model of determining authenticity of a human biometric image included in a received image dataset, by setting a parameter calculated for a first layer of the first neural network in the first layer included in a second neural network that includes the first layer and does not include the second layer.
8 . A non-transitory computer-readable storage medium storing a computer program that causes a computer to perform a process comprising:
calculating a parameter for each of a plurality of layers included in a first neural network through first machine learning using a plurality of image datasets each containing a human biometric image, the plurality of layers including a first layer and a second layer; and
generating a determination model of determining authenticity of a human biometric image included in a received image dataset, by setting a parameter calculated for the first layer of the first neural network in the first layer included in a second neural network that includes the first layer and does not include the second layer.