Systems and methods for predicting video quality based on objectives of video producer
Systems, methods, and non-transitory computer-readable media can collect a set of training videos as training data, wherein the set of training videos are labeled with one or more labels based on one or more video quality metrics associated with an evaluation objective. A machine learning model is trained based on the training data. A video to be evaluated is received. The video is assigned to a first video quality category of a plurality of video quality categories based on the machine learning model.
1. A computer-implemented method comprising:
collecting, by a computing system, a set of training videos as training data, wherein the set of training videos are labeled with one or more labels based on one or more video quality metrics associated with an evaluation objective;
training, by the computing system, a machine learning model based on the training data to determine a plurality of video quality categories, wherein each video quality category is associated with a respective time range;
receiving, by the computing system, a video to be evaluated; and
assigning, by the computing system, the video to a first video quality category of the plurality of video quality categories based on the machine learning model, wherein each video quality category of the plurality of video quality categories corresponds with a respective time range.
2. The computer-implemented method of claim 1 , wherein the video quality metric pertains to viewer retention time.
3. The computer-implemented method of claim 1 , wherein collecting the set of training videos comprises:
collecting a first set of training videos from a first set of pages of a social networking system.
4. The computer-implemented method of claim 3 , wherein collecting the set of training videos further comprises:
identifying a second set of pages of the social networking system that are similar to the first set of pages, and
collecting a second set of training videos from the second set of pages.
5. The computer-implemented method of claim 4 , wherein the second set of pages are identified based on a second machine learning model.
6. The computer-implemented method of claim 1 , wherein the machine learning model is a multi-stage model comprising a deep neural network and a sparse neural network.
7. The computer-implemented method of claim 6 , wherein the deep neural network is configured to receive image and sound data associated with the video, and generate a vector representation of the video.
8. The computer-implemented method of claim 7 , wherein the sparse neural network model is configured to receive metadata associated with the video and the vector representation of the video generated by the deep neural network, and generate respective likelihood scores corresponding to each of the plurality of video quality categories.
9. The computer-implemented method of claim 8 , wherein the video is assigned to the first video quality category of the plurality of video quality categories based on the first video quality category having a highest likelihood score of the plurality of video quality categories.
10. The computer-implemented method of claim 1 , wherein collecting the set of training videos comprises filtering out one or more videos from the set of training videos based on filtering criteria, wherein the filtering criteria comprise a minimum length threshold.
11. A system comprising:
at least one processor; and
a memory storing instructions that, when executed by the at least one processor, cause the system to perform a method comprising:
collecting a set of training videos as training data, wherein the set of training videos are labeled with one or more labels based on one or more video quality metrics associated with an evaluation objective;
training a machine learning model based on the training data to determine a plurality of video quality categories, wherein each video quality category is associated with a respective time range;
receiving a video to be evaluated; and
assigning the video to a first video quality category of the plurality of video quality categories based on the machine learning model, wherein each video quality category of the plurality of video quality categories corresponds with a respective time range.
12. The system of claim 11 , wherein the video quality metric pertains to viewer retention time.
13. The system of claim 11 , wherein collecting the set of training videos comprises: collecting a first set of training videos from a first set of pages of a social networking system.
14. The system of claim 13 , wherein collecting the set of training videos further comprises:
identifying a second set of pages of the social networking system that are similar to the first set of pages, and
collecting a second set of training videos from the second set of pages.
15. The system of claim 14 , wherein the second set of pages are identified based on a second machine learning model.
16. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to perform a method comprising:
collecting a set of training videos as training data, wherein the set of training videos are labeled with one or more labels based on one or more video quality metrics associated with an evaluation objective;
training a machine learning model based on the training data to determine a plurality of video quality categories, wherein each video quality category is associated with a respective time range;
receiving a video to be evaluated; and
assigning the video to a first video quality category of the plurality of video quality categories based on the machine learning model, wherein each quality category of the plurality of video quality categories corresponds with a respective time range.
17. The non-transitory computer-readable storage medium of claim 16 , wherein the video quality metric pertains to viewer retention time.
18. The non-transitory computer-readable storage medium of claim 16 , wherein collecting the set of training videos comprises: collecting a first set of training videos from a first set of pages of a social networking system.
19. The non-transitory computer-readable storage medium of claim 18 , wherein collecting the set of training videos further comprises:
identifying a second set of pages of the social networking system that are similar to the first set of pages, and
collecting a second set of training videos from the second set of pages.
20. The non-transitory computer-readable storage medium of claim 19 , wherein the second set of pages are identified based on a second machine learning model.