AI GAN enabled media compression for optimized resource utilization

1 . A computer-based method of compressing media utilizing a generative adversarial network (GAN), the method comprising:

receiving one or more media assets and historical data from a knowledge corpus in accordance with an identified usage context;

identifying, by a convolutional neural network (CNN), one or more objects in the one or more media assets;

deriving a relevance score for each identified object based on the historical data and the identified usage context, wherein the relevance score indicates a level of significance of each object in evaluating a quality of a task;

creating a training data set for a GAN generator including one or more images of a first set of one or more objects that exceed a relevance score threshold, wherein creating the training data set for the GAN generator further comprises training the GAN generator by feeding the created training data set into the GAN generator, wherein at least one object that does not exceed the relevance score threshold is removed from the created training data set;

applying, by the GAN generator, one or more modifications to each object in the first set based on the relevance score of each object;

determining whether a discriminator of the GAN is able to identify each object in the first set modified by the GAN generator; and

in response to determining the GAN discriminator is able to identify each object in the first set modified by the GAN generator as real:

generating, by the GAN generator, one or more updated media assets including a second set of one or more objects that are identified by the GAN discriminator as real, wherein the trained GAN generator is applied to the one or more media assets in the knowledge corpus that are consistent with the usage context, and wherein the one or more updated media assets depict the second set of one or more objects at varying levels of compression based on the relevance score of each object in the second set of one or more objects.

2 . The computer-based method of claim 1 , further comprising:

in response to determining the GAN discriminator is not able to identify each object in the first set modified by the GAN generator as real, iterating, until the GAN discriminator is able to identify each object in the first set as real:

applying, by the GAN generator, one or more additional modifications to each object in the first set not identified as real based on the relevance score of each object.

3 . The computer-based method of claim 2 , further comprising:

adding the updated one or more media assets to the knowledge corpus.

4 . The computer-based method of claim 1 , wherein applying the one or more modifications further comprises:

executing one or more compression techniques on each object in the first set.

5 . The computer-based method of claim 4 , wherein a degree of compression applied to each object in the first set is inversely proportional to the relevance score of each object in the first set, wherein an object having a lower relevance score is more compressed than an object having a higher relevance score.

6 . The computer-based method of claim 4 , wherein at least one compression technique includes adapting a pixel density of at least one object in the first set consistent with the relevance score of the at least one object.

7 . A computer system, the computer system comprising:

one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more computer-readable tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, wherein the computer system is capable of performing a method comprising:

receiving one or more media assets and historical data from a knowledge corpus in accordance with an identified usage context;

identifying, by a convolutional neural network (CNN), one or more objects in the one or more media assets;

deriving a relevance score for each identified object based on the historical data and the identified usage context, wherein the relevance score indicates a level of significance of each object in evaluating a quality of a task;

creating a training data set for a generative adversarial network (GAN) generator including one or more images of a first set of one or more objects that exceed a relevance score threshold, wherein creating the training data set for the GAN generator further comprises training the GAN generator by feeding the created training data set into the GAN generator, wherein at least one object that does not exceed the relevance score threshold is removed from the created training data set;

applying, by the GAN generator, one or more modifications to each object in the first set based on the relevance score of each object;

determining whether a discriminator of the GAN is able to identify each object in the first set modified by the GAN generator; and

in response to determining the GAN discriminator is able to identify each object in the first set modified by the GAN generator as real:

generating, by the GAN generator, one or more updated media assets including a second set of one or more objects that are identified by the GAN discriminator as real, wherein the trained GAN generator is applied to the one or more media assets in the knowledge corpus that are consistent with the usage context, and wherein the one or more updated media assets depict the second set of one or more objects at varying levels of compression based on the relevance score of each object in the second set of one or more objects.

8 . The computer system of claim 7 , the method further comprising:

in response to determining the GAN discriminator is not able to identify each object in the first set modified by the GAN generator as real, iterating, until the GAN discriminator is able to identify each object in the first set as real:

applying, by the GAN generator, one or more additional modifications to each object in the first set not identified as real based on the relevance score of each object.

9 . The computer system of claim 8 , the method further comprising:

adding the updated one or more media assets to the knowledge corpus.

10 . The computer system of claim 7 , wherein applying the one or more modifications further comprises:

executing one or more compression techniques on each object in the first set.

11 . The computer system of claim 10 , wherein a degree of compression applied to each object in the first set is inversely proportional to the relevance score of each object in the first set, wherein an object having a lower relevance score is more compressed than an object having a higher relevance score.

12 . The computer system of claim 10 , wherein at least one compression technique includes adapting a pixel density of at least one object in the first set consistent with the relevance score of the at least one object.

13 . A computer program product, the computer program product comprising:

one or more computer-readable tangible storage medium and program instructions stored on at least one of the one or more computer-readable tangible storage medium, the program instructions executable by a processor capable of performing a method, the method comprising:

receiving one or more media assets and historical data from a knowledge corpus in accordance with an identified usage context;

identifying, by a convolutional neural network (CNN), one or more objects in the one or more media assets;

deriving a relevance score for each identified object based on the historical data and the identified usage context, wherein the relevance score indicates a level of significance of each object in evaluating a quality of a task;

creating a training data set for a generative adversarial network (GAN) generator including one or more images of a first set of one or more objects that exceed a relevance score threshold, wherein creating the training data set for the GAN generator further comprises training the GAN generator by feeding the created training data set into the GAN generator, wherein at least one object that does not exceed the relevance score threshold is removed from the created training data set;

applying, by the GAN generator, one or more modifications to each object in the first set based on the relevance score of each object;

determining whether a discriminator of the GAN is able to identify each object in the first set modified by the GAN generator; and

in response to determining the GAN discriminator is able to identify each object in the first set modified by the GAN generator as real:

generating, by the GAN generator, one or more updated media assets including a second set of one or more objects that are identified by the GAN discriminator as real, wherein the trained GAN generator is applied to the one or more media assets in the knowledge corpus that are consistent with the usage context, and wherein the one or more updated media assets depict the second set of one or more objects at varying levels of compression based on the relevance score of each object in the second set of one or more objects.

14 . The computer program product of claim 13 , the method further comprising:

in response to determining the GAN discriminator is not able to identify each object in the first set modified by the GAN generator as real, iterating, until the GAN discriminator is able to identify each object in the first set as real:

applying, by the GAN generator, one or more additional modifications to each object in the first set not identified as real based on the relevance score of each object.

15 . The computer program product of claim 14 , the method further comprising:

adding the updated one or more media assets to the knowledge corpus.

16 . The computer program product of claim 13 , wherein applying the one or more modifications further comprises:

executing one or more compression techniques on each object in the first set.

17 . The computer program product of claim 16 , wherein a degree of compression applied to each object in the first set is inversely proportional to the relevance score of each object in the first set, wherein an object having a lower relevance score is more compressed than an object having a higher relevance score.