Method and apparatus for training neural network for generating deformed face image from face image, and storage medium storing instructions to perform method for generating deformed face image from face image
There is provided a method for training a neural network for generating a deformed face image from a face image preformed by an apparatus including a memory and a processor. The method comprises acquiring a face image for training; extracting facial feature information from the face image for training; and training the neural network to generate a deformed face image on the basis of the facial feature information, wherein the training includes generating multiple facial feature information that is gradually changed depending on the dosage or use duration of drugs, using the facial feature information as input data for training.
1 . An apparatus for training a neural network for generating a deformed face image from a face image, the apparatus comprising:
a storage medium configured to store one or more instructions;
an acquisition unit configured to acquire a face image for training; and
a processor configured to execute the one or more instructions stored in the storage medium, wherein the one or more instructions, when executed by the processor, cause the processor to extract a facial feature information from the face image for training, and train the neural network to generate the deformed face image, on a basis of the facial feature information,
wherein the processor is configured to generate multiple facial feature information that is changed depending on a dosage or use duration of drugs, on the basis of the facial feature information.
2 . The apparatus of claim 1 ,
wherein the multiple facial feature information includes at least one level information that is based on an amount of face image change, and
wherein the processor is configured to provide an output of preceding multiple facial feature information of the at least one level information as an input of following multiple facial feature information of the at least one level information.
3 . The apparatus of claim 2 ,
wherein the processor is configured to decode the multiple facial feature information to output multiple face images for each level.
4 . The apparatus of claim 1 ,
wherein the neural network includes a generator and a discriminator in a generative adversarial network, and
wherein the processor is configured to train the generator to generate a fake face image multiplexed for each level when the face image for training is input, and train the discriminator to classify the fake face image when the fake face image generated by the generator.
5 . The apparatus of claim 4 ,
wherein the processor is configured to receive a discrimination result output from the discriminator and train the generator to generate the fake face image.
6 . The apparatus of claim 1 ,
wherein the acquisition unit is configured to acquire face images of a plurality of drug criminals, and classify a deformation degree of each face image of the plurality of drug criminals into levels.
7 . The apparatus of claim 6 ,
wherein the acquisition unit is configured to determine the face image for training for extracting the facial feature information based on a priority or average value of results obtained by classifying the deformation degree into levels.
8 . The apparatus of claim 1 ,
wherein the acquisition unit is configured to convert the face image for training into a near infrared ray face image for training, and
wherein the processor is configured to extract facial feature information from the near infrared ray face image for training.