Evaluating visual quality of digital content
Systems, devices, methods, and computer readable medium for evaluating visual quality of digital content are disclosed. Methods can include training machine learning models on images. A request is received to evaluate quality of an image included in a current version of a digital component generated by the computing device. The machine learning models are deployed on the image to generate a score for each quality characteristic of the image. A weight is assigned to each score to generate weighted scores. The weighted scores are combined to generate a combined score for the image. The combined score is compared to one or more thresholds to generate a quality of the image.
1. A method comprising:
training, by a content distribution system comprising one or more processors, a plurality of machine learning models on a plurality of images, each image being assigned scores indicating quality characteristics of the image, each machine learning model being trained to score an image for a respective quality characteristic;
receiving, by the content distribution system from a computing device, a request to evaluate image quality of an image included in a current version of a digital component generated by the computing device, wherein the digital component includes the image and at least one other content asset that differs from the image;
deploying, by the content distribution system, the plurality of machine learning models on the image to generate multiple scores including a score for each quality characteristic;
assigning, by the content distribution system, a weight to each score among the multiple scores to generate weighted scores;
combining, by the content distribution system, the weighted scores to generate a combined quality score of the image; and
combining the combined quality score of the image with a quality score of the at least one other content asset included in a visual presentation of the digital component to generate an aggregate quality of the digital component that visually presents the image and the at least one other content asset that differs from the image.
2. The method of claim 1 , wherein:
the plurality of machine learning models comprise two or more of a blurriness model, an objectionable content model, or an orientation model;
the blurriness model generates a blurriness score quantifying blurriness within the image;
the objectionable content model generates a objectionable content score quantifying objectionable content within the image; and
the orientation model generates an orientation score indicating an orientation of the image.
3. The method of claim 1 , wherein the one or more machine learning models comprise a neural network model comprising a plurality of nodes and a plurality of layers, wherein at least one of a number of the nodes or a number of the layers is varied as the training of the neural network progresses so as to experimentally determine an ideal number of at least one of the nodes or the layers.
4. The method of claim 1 , further comprising:
aggregating, by the content distribution system, the combined score with other combined scores of other images in the digital component to generate an aggregate score for the digital component; and
updating, by the one or more processors, a graphical user interface of a first computing device to present the aggregate score.
5. The method of claim 4 , further comprising:
receiving, by the one or more processors, a modification of the image;
evaluating, by the one or more processors, the combined score for the modified image;
modifying, by the one or more processors, the aggregate score based on the combined score; and
updating, by the one or more processors, the graphical user interface of the first computing device to present the modified aggregate score.
6. The method of claim 4 , further comprising:
comparing, by the content distribution system, the aggregate score with preset quality heuristics;
determining, by the content distribution system, that the aggregate score does not comply with the preset quality heuristics;
generating, by the content distribution system in response to determining that the aggregate quality does not comply with the preset quality heuristics, one or more recommendations for improving the aggregate score; and
updating, by the one or more processors, the graphical user interface of the first computing device to present the one or more recommendations.
7. The method of claim 4 , further comprising:
comparing, by the content distribution system, the aggregate score with a threshold value;
determining, by the content distribution system, that the aggregate score is less than the threshold value; and
preventing, by the content distribution system in response to the determining that the aggregate score is less than the threshold value, distribution of the digital component to one or more client devices.
8. A system comprising:
a memory storing computer executable instructions; and
one or more computers configured to execute the instructions, wherein execution of the instructions cause the one or more computers to perform operations comprising:
training a plurality of machine learning models on a plurality of images, each image being assigned scores indicating quality characteristics of the image, each machine learning model being trained to score an image for a respective quality characteristic;
receiving, from a computing device, a request to evaluate image quality of an image included in a current version of a digital component generated by the computing device, wherein the digital component includes the image and at least one other content asset that differs from the image;
deploying the plurality of machine learning models on the image to generate multiple scores including a score for each quality characteristic;
assigning a weight to each score among the multiple scores to generate weighted scores;
combining the weighted scores to generate a combined quality score of the image;
combining the combined quality score of the image with a quality score of the at least one other content asset included in a visual presentation of the digital component to generate an aggregate quality of the digital component that visually presents the image and the at least one other content asset that differs from the image.
9. The system of claim 8 , wherein:
the plurality of machine learning models comprise two or more of a blurriness model, an objectionable content model, or an orientation model;
the blurriness model generates a blurriness score quantifying blurriness within the image;
the objectionable content model generates a objectionable content score quantifying objectionable content within the image; and
the orientation model generates an orientation score indicating an orientation of the image.
10. The system of claim 8 , wherein the one or more machine learning models comprise a neural network model comprising a plurality of nodes and a plurality of layers, wherein at least one of a number of the nodes or a number of the layers is varied as the training of the neural network progresses so as to experimentally determine an ideal number of at least one of the nodes or the layers.
11. The system of claim 8 , wherein the instructions cause the one or more computers to perform operations further comprising:
aggregating the combined score with other combined scores of other images in the digital component to generate an aggregate score for the digital component; and
updating a graphical user interface of a first computing device to present the aggregate score.
12. The system of claim 11 , wherein the instructions cause the one or more computers to perform operations further comprising:
receiving a modification of the image;
evaluating the combined score for the modified image;
modifying the aggregate score based on the combined score; and
updating the graphical user interface of the first computing device to present the modified aggregate score.
13. The system of claim 11 , wherein the instructions cause the one or more computers to perform operations further comprising:
comparing the aggregate score with preset quality heuristics;
determining that the aggregate score does not comply with the preset quality heuristics;
generating, in response to determining that the aggregate quality does not comply with the preset quality heuristics, one or more recommendations for improving the aggregate score; and
updating the graphical user interface of the first computing device to present the one or more recommendations.
14. A non-transitory computer readable medium storing instructions, that when executed by one or more computers, cause the one or more computers to perform operations comprising:
training a plurality of machine learning models on a plurality of images, each image being assigned scores indicating quality characteristics of the image, each machine learning model being trained to score an image for a respective quality characteristic;
receiving, from a computing device, a request to evaluate image quality of an image included in a current version of a digital component generated by the computing device, wherein the digital component includes the image and at least one other content asset that differs from the image;
deploying the plurality of machine learning models on the image to generate multiple scores including a score for each quality characteristic;
assigning a weight to each score among the multiple scores to generate weighted scores;
combining the weighted scores to generate a combined quality score of the image;
combining the combined quality score of the image with a quality score of the at least one other content asset included in a visual presentation of the digital component to generate an aggregate quality of the digital component that visually presents the image and the at least one other content asset that differs from the image.
15. The non-transitory computer readable medium of claim 14 , wherein:
the plurality of machine learning models comprise two or more of a blurriness model, an objectionable content model, or an orientation model;
the blurriness model generates a blurriness score quantifying blurriness within the image;
the objectionable content model generates a objectionable content score quantifying objectionable content within the image; and
the orientation model generates an orientation score indicating an orientation of the image.
16. The non-transitory computer readable medium of claim 14 , wherein the one or more machine learning models comprise a neural network model comprising a plurality of nodes and a plurality of layers, wherein at least one of a number of the nodes or a number of the layers is varied as the training of the neural network progresses so as to experimentally determine an ideal number of at least one of the nodes or the layers.
17. The non-transitory computer readable medium of claim 14 , wherein the instructions cause the one or more computers to perform operations further comprising:
aggregating the combined score with other combined scores of other images in the digital component to generate an aggregate score for the digital component; and
updating a graphical user interface of a first computing device to present the aggregate score.
18. The non-transitory computer readable medium of claim 17 , wherein the instructions cause the one or more computers to perform operations further comprising:
receiving a modification of the image;
evaluating the combined score for the modified image;
modifying the aggregate score based on the combined score; and
updating the graphical user interface of the first computing device to present the modified aggregate score.
19. The non-transitory computer readable medium of claim 17 , wherein the instructions cause the one or more computers to perform operations further comprising:
comparing the aggregate score with preset quality heuristics;
determining that the aggregate score does not comply with the preset quality heuristics;
generating in response to determining that the aggregate quality does not comply with the preset quality heuristics, one or more recommendations for improving the aggregate score; and
updating the graphical user interface of the first computing device to present the one or more recommendations.
20. The non-transitory computer readable medium of claim 17 , wherein the instructions cause the one or more computers to perform operations further comprising:
comparing the aggregate score with a threshold value;
determining that the aggregate score is less than the threshold value; and
preventing, in response to the determining that the aggregate score is less than the threshold value, distribution of the digital component to one or more client devices.