Automated visual suggestion, generation, and assessment using computer vision detection

1. A non-transitory computer-readable medium storing instructions which, when executed by a processor, cause the processor to:

receive a request comprising an audience definition of an audience;

identify a set of users based on the audience definition;

identify one or more images interacted with by the set of users;

determine a first image concept conveyed in the one or more images;

determine a first probability that the first image concept was at least a first reason that the one or more images were interacted with by at least some members of the audience;

generate an output based on the first probability; and

provide the output responsive to the request.

2. The non-transitory computer-readable medium of claim 1 , wherein the output is an image selected from the one or more images based on the first probability.

3. The non-transitory computer-readable medium of claim 1 , wherein the processor is further caused to:

determine a second image concept conveyed in the one or more images;

determine a second probability that the second image concept was at least a second reason that the one or more images were interacted with by at least some members of the audience; and

provide a ranked listing of the first image concept and the second image concept based on the first probability and the second probability.

4. The non-transitory computer-readable medium of claim 1 , wherein the processor is further caused to:

generate an image concept model that correlates attributes of users with image concepts determined from images interacted with by the set of users;

apply one or more user attributes in the audience definition to the image concept model; and

obtain the first image concept as an output of the image concept model.

5. The non-transitory computer-readable medium of claim 4 , wherein the processor is further caused to:

receive a candidate image to be provided to the audience, the candidate image based on the first image concept;

provide the candidate image to the audience;

determine a number of user interactions with the candidate image; and

update the image concept model based on the number of user interactions.

6. The non-transitory computer-readable medium of claim 4 , wherein the audience definition is received in association with a campaign, and wherein the processor is further caused to:

monitor a result of the campaign; and

update the image concept model based on the result.

7. The non-transitory computer-readable medium of claim 1 , wherein the processor is further caused to:

obtain a plurality of images interacted with by online-system users; and

for each image of the plurality of images:

identify at least one image concept of the image based on computer vision analysis; and

store, in an image concept store, the at least one image concept in association with the image, wherein the first image concept is determined based on the image concept store.

8. The non-transitory computer-readable medium of claim 7 , wherein, to identify the at least one image concept of the image, the processor is further caused to:

identify a first location of the first image concept in the image;

identify a second location of a second image concept in the image; and

determine a location context based on the first location and the second location, wherein the generated output is further based on the location context.

9. The non-transitory computer-readable medium of claim 1 , wherein to generate the output, the processor is further caused to:

generate an image concept recommendation comprising an identification of the first image concept to be included in an image to be provided to the audience.

10. The non-transitory computer-readable medium of claim 9 , wherein the processor is further caused to:

identify a sample image that includes the first image concept; and

include the sample image with the image concept recommendation.

11. The non-transitory computer-readable medium of claim 1 , wherein to generate the output, the processor is further caused to:

generate a candidate image based on the first image concept.

12. The non-transitory computer-readable medium of claim 11 , wherein to generate the candidate image, the processor is further caused to:

identify the audience based on the audience definition, the audience to receive the candidate image; and

include the first image concept in the candidate image based on a prediction that the first image concept appeals to the audience.

13. The non-transitory computer-readable medium of claim 1 , wherein to generate the output, the processor is further caused to:

receive a candidate image to be provided to the audience;

identify one or more image concepts from the candidate image; and

determine an assessment that comprises a probability that the candidate image has a likelihood of being interacted with by the audience based on the one or more image concepts identified from the candidate image and the audience definition.

14. A non-transitory computer-readable medium having instructions which, when executed by a processor, cause the processor to:

process a plurality of images, wherein for each image, the processing comprises:

identifying one or more image concepts from the image based on computer vision analysis of the image;

storing, in an image store, the one or more image concepts in association with the image;

providing the image for interaction to a plurality of users of an online system;

determining a set of users, from among the plurality of users, that interacted with the image; and

storing, in an action log, an indication of the interaction;

receive a request comprising an audience definition of an audience;

identify the set of users based on the audience definition;

identify one or more images interacted with by the set of users based on the action log;

determine an image concept based on the one or more images; and

generate an output based on the image concept.

15. The non-transitory computer-readable medium of claim 14 , wherein the processor is further caused to:

generate an image concept recommendation comprising an identification of the image concept to be included in an image provided to the audience.

16. The non-transitory computer-readable medium of claim 15 , wherein the processor is further caused to:

identify a sample image that includes the image concept; and

include the sample image with the image concept recommendation.

17. The non-transitory computer-readable medium of claim 14 , wherein the processor is further caused to:

generate a candidate image based on the image concept.

18. The non-transitory computer-readable medium of claim 14 , wherein the processor is further caused to:

receive a candidate image to be provided to the audience;

identify one or more image concepts from the candidate image; and

determine an assessment that comprises a probability that the candidate image has a likelihood of being interacted with by the audience based on the one or more image concepts identified from the candidate image and the audience definition.

19. A non-transitory computer-readable medium having instructions which, when executed by a processor, cause the processor to:

identify interactions with a plurality images by a user, wherein a plurality of attributes is known about the user;

identify a plurality of image concepts from the plurality of images;

determine a correlation between an attribute of the user with one or more of the plurality of image concepts; and

generate an image concept model based on the determined correlation.

20. The non-transitory computer-readable medium of claim 19 , wherein the processor is further caused to:

generate, via the image concept model, a recommendation to use an image concept in a candidate image;

monitor user interactions with the candidate image; and

update the image concept model based on the monitored user interactions.