No reference realtime video quality assessment
A method for video quality assessment implemented by a computing device, where the method includes selecting a first set of frames from an input video source, determining features of the first set of frames for input into a first machine learning model, applying the first machine learning model to the features of the first set of frames to obtain a video quality score and a confidence score, and triggering a retraining of the first machine learning model in response to the confidence score being determined to be out of bounds.
1 . A method implemented by a computing device, the method comprising:
selecting a set of frames for video quality assessment on a first input video source, the set of frames to be distributed from the first input video source to a content delivery network;
determining a set of features associated with the set of frames for a reference machine learning model;
applying the reference machine learning model to the set of features associated with the set of frames to obtain a video quality score and a confidence score that indicates a confidence level of the video quality score;
training a set of machine learning models using the set of features, the video quality score, and the confidence score, a machine learning model within the set of machine learning models trained to be applicable to a second input video source; and
checking for further frames from the first input video source to trigger further frame selection and feature determination for the reference machine learning model in response to existence of the further frames and to terminate the video quality assessment on the first input video source otherwise.
2 . The method of claim 1 , wherein training the set of machine learning model comprises training a first machine learning model using a first set of features within the set of features and training a second machine learning model using a second set of features within the set of features that are different from the first set of features.
3 . The method of claim 1 , wherein the second input video source is categorized to be in a same type of events as that of the first input video source.
4 . The method of claim 1 , wherein the set of frames are preprocessed to enhance image quality prior to applying the reference machine learning model.
5 . The method of claim 1 , wherein the set of frames are selected in a streaming platform that receives the set of frames from the first input video source as a broadcasting live video feed.
6 . The method of claim 1 , wherein the set of frames are selected in a cloud region of a cloud computing environment prior to forwarding to the content delivery network.
7 . The method of claim 1 , wherein the set of features includes one or more of quantization parameters, motion vectors, bit rate, scene statistics, standard deviation of quantization parameters, and standard deviation of motion vector.
8 . An electronic device comprising:
a non-transitory machine-readable storage medium to store instructions; and
a processor coupled to the non-transitory machine-readable storage medium, wherein the instructions, when executed by a processor, are capable of causing the processor to perform operations that are to:
select a set of frames for video quality assessment on a first input video source, the set of frames to be distributed from the first input video source to a content delivery network,
determine a set of features associated with the set of frames for a reference machine learning model,
apply the reference machine learning model to the set of features associated with the set of frames to obtain a video quality score and a confidence score that indicates a confidence level of the video quality score,
train a set of machine learning models using the set of features, the video quality score, and the confidence score, a machine learning model within the set of machine learning models trained to be applicable to a second input video source, and
check for further frames from the first input video source to trigger further frame selection and feature determination for the reference machine learning model in response to existence of the further frames and to terminate the video quality assessment on the first input video source otherwise.
9 . The electronic device of claim 8 , wherein training the set of machine learning model comprises training a first machine learning model using a first set of features within the set of features and training a second machine learning model using a second set of features within the set of features that are different from the first set of features.
10 . The electronic device of claim 8 , wherein the second input video source is categorized to be in a same type of events as that of the first input video source.
11 . The electronic device of claim 8 , wherein the set of frames are preprocessed to enhance image quality prior to applying the reference machine learning model.
12 . The electronic device of claim 8 , wherein the set of frames are selected in a streaming platform that receives the set of frames from the first input video source as a broadcasting live video feed.
13 . The electronic device of claim 8 , wherein the set of frames are selected in a cloud region of a cloud computing environment prior to forwarding to the content delivery network.
14 . The electronic device of claim 8 , wherein the set of features includes one or more of quantization parameters, motion vectors, bit rate, scene statistics, standard deviation of quantization parameters, and standard deviation of motion vector.
15 . A non-transitory machine-readable storage medium that provides instructions that, when executed by a processor, are capable of causing the processor to perform operations, the operations comprising:
selecting a set of frames for video quality assessment on a first input video source, the set of frames to be distributed from the first input video source to a content delivery network;
determining a set of features associated with the set of frames for a reference machine learning model;
applying the reference machine learning model to the set of features associated with the set of frames to obtain a video quality score and a confidence score that indicates a confidence level of the video quality score;
training a set of machine learning models using the set of features, the video quality score, and the confidence score, a machine learning model within the set of machine learning models trained to be applicable to a second input video source; and
checking for further frames from the first input video source to trigger further frame selection and feature determination for the reference machine learning model in response to existence of the further frames and to terminate the video quality assessment on the first input video source otherwise.
16 . The non-transitory machine-readable storage medium of claim 15 , wherein training the set of machine learning model comprises training a first machine learning model using a first set of features within the set of features and training a second machine learning model using a second set of features within the set of features that are different from the first set of features.
17 . The non-transitory machine-readable storage medium of claim 15 , wherein the second input video source is categorized to be in a same type of events as that of the first input video source.
18 . The non-transitory machine-readable storage medium of claim 15 , wherein the set of frames are preprocessed to enhance image quality prior to applying the reference machine learning model.
19 . The non-transitory machine-readable storage medium of claim 15 , wherein the set of frames are selected in a streaming platform that receives the set of frames from the first input video source as a broadcasting live video feed.
20 . The non-transitory machine-readable storage medium of claim 15 , wherein the set of frames are selected in a cloud region of a cloud computing environment prior to forwarding to the content delivery network.