WO2023185175A1

WO2023185175A1 - Video processing method and apparatus

Info

Publication number: WO2023185175A1
Application number: PCT/CN2022/144210
Authority: WO
Inventors: 侯芬; 何钧; 张希文
Original assignee: 上海哔哩哔哩科技有限公司
Priority date: 2022-03-28
Filing date: 2022-12-30
Publication date: 2023-10-05
Also published as: CN114693812A

Abstract

The present application provides a video processing method and apparatus. The method comprises: when a target video is not played back, inputting an object feature of a target object corresponding to the target video to an object feature decision model to obtain a first video decision result; when the first video decision result does not meet a decision condition, playing back the target video; within a target playback time period, determining a second video decision result according to a preset video processing strategy; and when the second video decision result meets the decision condition, performing video processing on the target video. According to the method, a transcoding decision model and a transcoding decision strategy are designed, a video transcoding decision is made in advance by means of the transcoding decision model before a video is opened, the decision is supplemented in a timely a manner by means of the transcoding decision strategy after the video is opened, thereby achieving timely and effective video transcoding.

Description

Video processing method and device

This application declares the priority of the Chinese patent application with application number 202210311587.2 and titled "Video Processing Method and Device" submitted on March 28, 2022. The entire content of the Chinese patent application is incorporated into this application by reference.

Technical field

The present application relates to the field of video processing technology, and in particular to a video processing method. This application also relates to a video processing device, a computing device, and a computer-readable storage medium.

Background technique

Video coding is an important technical means in the video field. Unencoded videos may be larger in size, which will put great pressure on the storage and transmission of the videos. Therefore, when storing and transmitting videos, video data is generally compressed through video coding.

However, the inventor realized that the currently adopted strategies for deciding whether to encode a video, that is, transcoding, are relatively simple. One is to transcode all videos indiscriminately, and the other is to transcode through human subjective judgment. Coding decision-making; however, both strategies will lead to waste of resources or insufficient coding.

Contents of the invention

In view of this, embodiments of the present application provide a video processing method. The present application also relates to a video processing device, a computing device, and a computer-readable storage medium to solve the technical problems of resource waste or insufficient coding during video transcoding that exist in the prior art.

According to a first aspect of the embodiments of the present application, a video processing method is provided, including:

When the target video is not played, input the object characteristics of the target object corresponding to the target video into the object characteristics decision model to obtain the first video decision result;

When the first video decision result does not meet the decision conditions, play the target video;

Within the target playback time period, determine the second video decision result according to the preset video processing strategy;

If the second video decision result satisfies the decision condition, perform video processing on the target video.

According to a second aspect of the embodiment of the present application, a video processing device is provided, including:

The first result obtaining module is configured to input the object characteristics of the target object corresponding to the target video into the object characteristic decision model to obtain the first video decision result when the target video is not played;

A video playback module configured to play the target video when the first video decision result does not meet the decision condition;

The second result obtaining module is configured to determine the second video decision result according to the preset video processing strategy within the target playback time period;

A video processing module configured to perform video processing on the target video if the second video decision result satisfies the decision condition.

According to a third aspect of the embodiment of the present application, a computing device is provided, including a memory, a processor, and computer instructions stored in the memory and executable on the processor. When the processor executes the instructions, the described instructions are implemented. Steps of video processing methods.

According to a fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, which stores computer instructions that implement the steps of the video processing method when executed by a processor.

The video processing method provided by this application includes, when the target video is not played, inputting the object characteristics of the target object corresponding to the target video into the object characteristics decision model to obtain the first video decision result; in the first video When the decision result does not meet the decision condition, the target video is played; within the target play time period, the second video decision result is determined according to the preset video processing strategy; when the second video decision result meets the decision condition In this case, perform video processing on the target video.

Specifically, before the target video is played, the video processing method uses the object characteristics of the target object corresponding to the target video and combines the pre-trained object characteristics decision model to make the video transcoding decision in advance; and the video transcoding decision is When the video is not transcoded, after the target video is played, the video transcoding decision is made again through the preset video processing strategy to solve the waste of resources caused by transcoding all target videos, and after determining the video to be transcoded based on the above strategy. In the case of encoding, the target video is effectively transcoded to avoid insufficient encoding and achieve accurate transcoding of the target video.

Description of drawings

Figure 1 is an exemplary illustration of a video processing method provided by an embodiment of the present application in a specific application scenario;

Figure 2 is a flow chart of a video processing method provided by an embodiment of the present application;

Figure 3 is a processing flow chart of a video processing method applied to a video transcoding scenario provided by an embodiment of the present application;

Figure 4 is a schematic structural diagram of a video processing device provided by an embodiment of the present application;

Figure 5 is a structural block diagram of a computing device provided by an embodiment of the present application.

Detailed ways

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be implemented in many other ways different from those described here. Those skilled in the art can make similar extensions without violating the connotation of the present application. Therefore, the present application is not limited by the specific implementation disclosed below.

The terminology used in one or more embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to limit the one or more embodiments of the present application. As used in one or more embodiments of this application and the appended claims, the singular forms "a," "the" and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It will also be understood that the term "and/or" as used in one or more embodiments of this application refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of the present application, the information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be called the second, and similarly the second may also be called the first. Depending on the context, the word "if" as used herein may be interpreted as "when" or "when" or "in response to determining."

First, the terminology involved in one or more embodiments of this application is explained.

Data mining: It is based on machine learning, pattern recognition, statistics, database and other disciplines to mine potential information from data to help decision-makers solve practical problems.

Machine learning: It is a discipline dedicated to how to use experience to improve the performance of the system itself through computational means. The main content of its research is about algorithms that generate "models" from data on computers, that is, "learning algorithms" .

Statistics: Statistics is the science of understanding the overall quantitative characteristics and quantitative relationships of objective phenomena. It is a methodological science that understands the quantitative regularity of objective phenomena through collecting, organizing, and analyzing statistical data.

Feature engineering: It is a means of extracting features from original data to the maximum extent for use by models and algorithms, including data preprocessing, feature selection, dimension and feature expansion, etc.

Video transcoding: Encode videos to achieve video data compression. Common encoding standards include JPEG, MJPEG, H264, H265, AV1, etc.

Transcoding decision: whether to transcode the video.

Bandwidth revenue: the reduction in bandwidth billing after video transcoding.

XGBoost model: It is a gradient boosting decision tree (GBDT, Gradient Boosting DecisionTree). XGBoost is essentially a method based on a tree structure and combined with integrated learning. Its basic tree structure is a classification and regression tree (CART, Classification and Regression Tree) ).

In this application, a video processing method is provided. This application also relates to a video processing device, a computing device, and a computer-readable storage medium, which will be described in detail one by one in the following embodiments.

Referring to Figure 1, Figure 1 is an exemplary illustration of a video processing method provided by an embodiment of the present application in a specific application scenario.

The specific application scenario in Figure 1 includes a client 102 and a server 104.

During specific implementation, the user (such as the uploader of the video) sends the video to be played to the server 104 through the client 102.

After receiving the video to be played, the server 104 determines the user of the video to be played, and obtains the proportion of videos with more than 10,000 views among all videos of the user. When the proportion of the video is greater than the preset proportion threshold Next, transcode the video to be played, and send the transcoded video to the client (including but not limited to client 102) for playback; wherein, the preset proportion threshold can be calculated based on historical data, as follows The specific calculation process will be introduced in the embodiment.

When the proportion of the video is less than the preset proportion threshold, obtain the user's data in multiple dimensions, such as the number of fans, the number of forwards, the number of likes, etc.; perform data processing on these multi-dimensional data to obtain User characteristics of this user. Input the user characteristics of the user into the pre-trained object characteristic decision model, obtain the label corresponding to the user characteristics, and determine the video transcoding decision result based on the label, that is, when the label is 0, the video transcoding decision result is not transcoding. code; when the tag is 1, the video transcoding decision result is transcoding; then when the video transcoding decision result is transcoding, the video to be played is transcoded, and the transcoded video Send to client for playback.

When the decision result of the video is not to transcode, a decision is made as to whether to transcode the video according to the preset decision strategy. Specifically, the video is played, and within a preset time period (for example, within 48 hours) after the video is played, a video segment is divided every three minutes or five minutes, and the playback volume of the next video segment is compared with the previous video. When the number of views on a clip increases to more than 1,000, the video will be transcoded and the video that has been played will be supplemented and transcoded; or the playback rate of each video clip can be calculated based on statistics Volume threshold, that is, the popularity threshold. When the play volume of three or four consecutive video clips exceeds the play volume threshold, the video will be transcoded, and the video that has been played will be supplemented and transcoded; And send the transcoded video to the client for playback.

If the preset time period ends after the video is played and the video is still not transcoded, the number of views, reposts, and likes after the video is played within the preset time period are obtained, and the playback volume, reposts, etc. of these videos are obtained. Perform data processing on the volume and number of likes, etc., to obtain the video playback characteristics after the video is played. Input the video playback features into the pre-trained playback feature decision model, obtain the label corresponding to the playback feature, and determine the video transcoding decision result based on the label. That is, if the label is 0, the video transcoding decision result is no transcoding. ; When the tag is 1, the video transcoding decision result is transcoding; then when the video transcoding decision result is transcoding, the video is transcoded, and the video that has been played is supplementary transcoded. And send the transcoded video to the client for playback.

The video processing method provided by the embodiments of this application uses machine learning and statistical methods to implement a transcoding decision-making scheme consisting of two models and a decision-making strategy sequence, achieving timely and effective transcoding decisions for videos. Specifically, before the video is played, the multi-dimensional data of the user to which the video belongs (such as the number of fans, the number of forwards, the number of likes, etc.) and the video information (such as the proportion of video playback volume exceeding 10,000, etc.) are used, combined with the first machine Learning model (i.e. object feature decision-making model), makes a video transcoding decision in advance to trigger transcoding; and when the video transcoding decision of the first machine learning model is not to transcode, the video is played and the video is played at the preset time Within (such as within 48 hours), use real-time popularity data (such as playback volume on video clips, etc.) combined with the preset transcoding decision strategy to make transcoding decisions; and no video transcoding is performed through this preset transcoding decision strategy. In this case, the data of a few days after the video is played (such as the number of video views, reposts, likes, etc.) combined with the playback feature decision model will be used to determine the video transcoding decision result. Due to the different stages of the video, the available data is different. The timeliness of video transcoding of the two models and a decision strategy decreases in sequence, while the accuracy of video transcoding increases in sequence. The two models and a decision strategy are executed sequentially. , making a decision on whether to transcode the video can not only achieve the timeliness of transcoding, but also achieve the accuracy of transcoding, so that the video can be transcoded in a timely and effective manner to maximize the difference between revenue and cost.

Referring to Figure 2, Figure 2 shows a flow chart of a video processing method according to an embodiment of the present application, which specifically includes the following steps:

Step 202: When the target video is not played, input the object characteristics of the target object corresponding to the target video into the object characteristic decision model to obtain the first video decision result.

The target video can be understood as the above-mentioned video to be played, which can be any type, any length, any format of video, such as sports videos, entertainment videos, two-hour movies, etc.

The target object corresponding to the target video can be understood as the uploader of the target video; the object characteristics of the target object can be understood as the object characteristics formed after data processing of the object attribute information of the target object, where the object attribute information includes but is not limited to Number of fans, number of retweets, number of likes, etc.

In actual applications, there will be some object attribute information that cannot be used directly. For example, a video partition has only one partition number. The partition number itself has no meaning, but the video partition itself will be a popular partition or an unpopular partition; if the partition number is used directly As an object feature, the video partition has no meaning. Therefore, some data processing will be performed on the video partition, and a score is configured for the video partition to indicate that the video partition is a popular partition or an unpopular partition, thus forming a realistic Object characteristics of meaning.

Therefore, after obtaining the object attribute information of the target object, the object attribute information will be data processed to determine the object characteristics of the target object. Subsequently, the object characteristics can be combined with the object characteristics decision-making model to quickly and accurately obtain the first Video decision results. The specific implementation method is as follows:

The method of inputting the object characteristics of the target object corresponding to the target video and the object characteristics decision-making model to obtain the first video decision-making result includes:

Obtain object attribute information of the target object corresponding to the target video;

Perform data processing on the object attribute information to obtain the object characteristics of the target object;

The object characteristics of the target object are input into the object characteristics decision model to obtain the first video decision result.

The detailed introduction of the object attribute information of the target object can be found in the above embodiments and will not be described again here; the object feature decision-making model includes but is not limited to the XGBoost model.

And perform data processing on the object attribute information to obtain the object characteristics of the target object; it can be understood that if some object attribute information is used directly as object characteristics, the use effect is not obvious, such as the partition number of the above-mentioned video partition, etc. Then the object attribute information of the video partition can be processed and processed, such as setting a metric value for the video partition that can distinguish hot and cold conditions, so that the object attribute information of the video partition becomes a meaningful object feature.

When used specifically, the object characteristics of the target object are input into the pre-trained object characteristic decision model to obtain the first video decision result of the target video; in actual applications, the specific application scenarios of this video processing method are different, and the first video decision The results will be different. For example, if the video processing method is applied to the video transcoding scenario, the first video decision result can be video transcoding or the video is not transcoded; if the video processing method is applied to the video fast-forwarding scenario, the first video decision result can be The video decision result can be fast forwarding of the video or not fast forwarding of the video.

For ease of understanding, the following embodiments take the video processing method being applied to a video transcoding scenario as an example. However, this does not limit the video processing method to other achievable scenarios, such as the video fast-forward scenario introduced above.

In addition, before using the object feature decision model to decide whether to transcode a video, the object feature decision model needs to be pre-trained to ensure the speed and accuracy of subsequent video transcoding decisions. The specific implementation method is as follows:

The training steps of the object feature decision-making model are as follows:

Obtain sample videos, determine the sample objects corresponding to each sample video, and the video playback volume;

Determine the training sample according to the object attribute information of the sample object;

Determine the sample label corresponding to the training sample according to the video playback volume;

The object feature decision model is trained according to the training samples and the sample labels.

Among them, the sample video can be understood as multiple videos of any type, any playing time, and any format that have been played in history; the sample object of the sample video can be understood as the uploader of the sample video; the video playback volume of the sample video can be understood The number of times the sample video has been played, that is, the number of times it has been viewed; the object attribute information of the sample object is the same as the object attribute information of the target object in the above embodiment, and can be understood as the number of fans, the number of forwards, the number of likes, etc.

During the specific implementation, multiple sample videos are first obtained, and then the sample object and video playback volume corresponding to each sample video are determined; the training sample is determined based on the object attribute information of the sample object, and the sample label corresponding to each training sample is determined based on the video playback volume. ; Then train the object feature decision-making model based on the training samples and training labels.

Among them, the sample label corresponding to the training sample is determined based on the video playback volume. It can be understood that when the video playback volume is greater than or equal to the preset playback volume threshold (such as 500), the sample label is set to 1, and the sample label corresponding to the sample label is determined. The training sample is a positive sample; when the video playback volume is less than the preset playback volume threshold, the sample label is set to 0, and the training sample corresponding to the sample label is determined to be a negative sample.

For example, if the video playback volume corresponding to a certain sample video is 1,000, then when the preset playback volume threshold is 500, it can be determined that the object attribute information of the sample object corresponding to the sample video is a training sample. For the corresponding video playback volume, the sample label determined for this training sample is 1, which means that the training sample is a positive sample.

In actual applications, there may be some invalid information in the object attribute information of the sample object. In order to improve the training effect of the object feature decision-making model, the object attribute information of the sample object will be data processed to achieve feature construction; so that subsequent Based on the constructed more reasonable object features as training samples, the object feature decision-making model is trained and the accuracy of the model is improved. The specific implementation method is as follows:

Determining the training sample based on the object attribute information of the sample object includes:

Perform data processing on the object attribute information of the sample object to obtain the object characteristics of the sample object;

The object characteristics of the sample object are determined as training samples.

The specific implementation method of performing data processing on the attribute information of the sample object and obtaining the object characteristics of the sample object can be found in the detailed introduction of the above embodiments, and will not be described again here.

Specifically, after obtaining the object attribute information of the sample object, data processing is performed on the attribute information of the sample object to obtain the object characteristics of the sample object; subsequently, the object characteristics of the sample object can be used as a training sample, and then combined with the video playback volume according to the The sample label of the training sample is determined to implement the training of the object feature decision-making model, so as to improve the use effect of the object feature decision-making model.

Before using the object feature decision-making model, if the target video corresponding to the target object has a relatively high playback volume in history, then the target video will have a high probability of being played in large numbers after it is uploaded. Therefore, the target video will not be played until the target video is uploaded. Before playing, you can first determine whether the video needs to be transcoded based on the historical video data of the target object corresponding to the target video, combined with the current playback status of other videos, to avoid the situation where the target video will be played in large numbers immediately after it goes online. The transcoding is not timely, which affects the viewing experience.

For example, from the above training samples, determine the positive sample video, then determine the historical data of the object of the positive sample video under a certain feature, and determine the feature threshold under the feature; for specific use, you can obtain the target video This feature of the target object is compared with the feature threshold obtained under the positive sample to determine whether the target video requires video transcoding. Specifically, the specific acquisition method of the feature threshold is as follows:

After determining the sample label corresponding to the training sample according to the video playback volume, the method further includes:

Determine the positive sample video in the sample video according to the sample label;

Determine the positive sample object corresponding to the positive sample video, and determine the target characteristics according to the historical video data of the positive sample object;

The corresponding feature threshold is determined according to the target feature.

Among them, the target feature can be any feature, such as the proportion of videos with more than 10,000 likes or the proportion of videos with more than 10,000 views; when the target feature is the proportion of videos with more than 10,000 likes, according to the positive sample object Determining the target characteristics of the historical video data can be understood as obtaining the number of likes for each video in all the videos uploaded by the positive sample object in the past, and determining the number of likes for all videos of the positive sample object based on the number of likes for each video. The proportion of videos with more than 10,000 likes, that is, the proportion of videos with more than 10,000 likes; when the target feature is the proportion of videos with more than 10,000 views, the target feature is determined based on the historical video data of the positive sample object, which can be understood as, Obtain the playback volume of each video among all the videos uploaded by the positive sample object in history, and determine the proportion of all videos of the positive sample object that have a playback volume of more than 10,000 based on the playback volume of each video, that is, the proportion of videos with a playback volume of more than 10,000 .

Specifically, taking the target feature as the proportion of video playback volume exceeding 10,000 as an example, determining the corresponding feature threshold based on the target feature can be understood as obtaining the proportion of video playback volume exceeding 10,000 for all positive sample objects, and determining its mean value. , minimum value, maximum value and other statistics; train the weight of each statistic, and finally the weighted sum of each statistic obtains the final threshold (i.e., feature threshold). In practical applications, the training of weights involves traversing the parameter combinations of weights and testing the final set of weight values on the test set. For example, to train the weights of the minimum and mean values, assume that the step size is 0.1. Then the parameter combinations include 0.1, 0.9; 0.2, 0.8; 0.3, 0.7, etc.

In the embodiment of this specification, the uploader of currently known transcoded videos obtains the proportion of videos with more than 10,000 views among all historical videos, or the proportion of videos with more than 10,000 likes among all historical videos; subsequently, based on each upload The proportion of each uploader's video playback volume exceeding 10,000 or the proportion of each uploader's video's likes volume exceeding 10,000 is calculated to calculate the corresponding feature threshold; when later determining whether the target video is transcoded, the target video can be first uploaded based on the corresponding By comparing the proportion of video views with over 10,000 views or the proportion of videos with over 10,000 likes and their corresponding feature thresholds, you can quickly determine whether the target video needs to be transcoded. The specific implementation method is as follows:

Before inputting the object characteristics of the target object into the object characteristic decision-making model and obtaining the first video decision-making result, the method further includes:

Determine the target feature of the target object corresponding to the target video, and the characteristic value of the target feature;

Obtain a fourth video decision result according to the correlation between the characteristic value of the target characteristic and the characteristic threshold;

When the fourth video decision result satisfies the decision condition, video processing is performed on the target video according to the fourth video decision result.

Among them, the target feature of the target object is the same as the target feature obtained by obtaining the feature threshold. For example, the feature threshold obtained above is the proportion of video playback volume exceeding 10,000, then the target feature is the proportion of video playback volume exceeding 10,000; and the target The characteristic value of a feature is a specific proportion value.

Specifically, when the target video is not played, the target feature of the target object corresponding to the target video and the characteristic value of the target feature are obtained, that is, the proportion and proportion of the video playback volume exceeding 10,000; The proportion of the proportion is compared with the characteristic threshold obtained above, and the fourth video decision result is obtained based on the specific comparison result, that is, the video is transcoded or not transcoded; finally, it is determined that the fourth video decision result is transcoded. In this case, it is determined that the fourth video decision result satisfies the decision condition; at this time, the target video can be transcoded according to the fourth video decision result.

In practical applications, the fourth video decision result is obtained based on the correlation between the characteristic value of the target feature and the characteristic threshold. It can be understood that the characteristic value of the target feature is compared with the characteristic threshold, and the target feature is If the feature value of is greater than or equal to the feature threshold, the fourth video decision result is determined to be video transcoding; if the feature value of the target feature is less than the feature threshold, the fourth video decision result is determined to be video non-transcoding.

For example, if the feature threshold is 70%, the proportion of videos with more than 10,000 videos is 75%, and the 75% proportion of videos with more than 10,000 videos is compared with the feature threshold of 70% obtained above, then It can be determined that the proportion of the video that exceeds 10,000 is greater than the feature threshold. At this time, it can be determined that the fourth video decision result is video transcoding.

In the embodiment of this specification, in order to ensure the timeliness of transcoding the target video, before transcoding the target video, features calculated based on the target characteristics of the target object corresponding to the target video and the target characteristics of historical transcoded videos can be used Compare the threshold and quickly decide whether to transcode the target video, so that when the target object's historical video playback volume exceeds 10,000 and accounts for a large proportion, the target video uploaded by the target object will be played in large quantities by default. The probability is relatively high, so transcoding and playback can be performed directly, improving the timeliness of transcoding and subsequent video playback revenue.

Step 204: If the first video decision result does not meet the decision condition, play the target video.

Specifically, the specific application scenarios of the video processing method provided by the embodiments of this specification are different, the first video decision result is different, and the content of judging whether the first video decision result meets the decision condition is also different; for example, the video processing method is applied to video transcoding In the scenario, the first video decision result can be understood as video transcoding or video non-transcoding introduced in the above embodiment, and the corresponding video decision conditions can be understood as video transcoding conditions; if the video processing method is applied to the video recommendation scenario , the first video decision result can be understood as video recommendation or video non-recommendation, and the corresponding video decision conditions can be understood as video recommendation conditions, etc.

Then, when the first video decision result is that the video is not transcoded, it can be determined that the first video decision result does not meet the decision condition; at this time, the original version of the target video is played without transcoding. video data.

When the first video decision result is video transcoding, it can be determined that the first video decision result satisfies the decision condition; at this time, the target video is directly transcoded and the transcoded target video is played.

Step 206: Within the target playback time period, determine the second video decision result according to the preset video processing strategy.

The target playback time period can be set according to the actual application, and the embodiment of the present application does not impose any limitation on this; for example, the target playback time period is set to 48 hours or 50 hours, etc.

Moreover, the preset video processing strategy can also be set according to the actual application. For example, the video can be divided into segments, and whether the video needs to be transcoded is determined based on the playback volume of the segmented video; it can also be based on all plays within the target playback time period. The overall number of views or likes of the video determines whether the video needs to be transcoded.

In the embodiment of this application, the video is divided into segments, and whether the video needs to be transcoded is determined based on the playback volume of the segmented video as a preset video processing strategy. Within the target playback time period, the preset video processing strategy is used to determine whether the video needs to be transcoded. Determine the decision-making results of the second video and introduce them in detail. The specific implementation method is as follows:

Determining the second video decision result according to the preset video processing strategy includes:

At least two video clips are obtained according to the preset division rules, and the second video decision result is determined based on the playback volume of the at least two video clips.

Among them, the preset division rules can be set according to actual applications, and the embodiments of this application again do not impose any limitations. For example, the preset division rule is to divide a video clip every three minutes, or divide a video clip every five minutes, etc.

For ease of understanding, the following embodiments take the preset dividing rule of dividing a video segment every three minutes as an example for detailed introduction.

In specific implementation, starting from the target video, one video segment is divided every three minutes. When at least two video segments are divided, the second video decision result is determined based on the playback volume of the at least two divided video segments.

The video processing method provided by the embodiment of the present application can obtain at least two video clips according to the preset division rules when the target video starts to be played. Subsequently, the third video clip can be quickly determined based on the playback volume of the divided at least two video clips. The second video decision result is used to ensure the timely transcoding of the target video.

After dividing at least two video clips, specific implementation methods for determining the second video decision result include at least two methods according to the playback volume of the divided at least two video clips. One method can combine two adjacent video clips. Compare the playback volume, and quickly determine the decision-making result of the second video based on the growth of playback volume. The specific implementation method is as follows:

Determining the second video decision result based on the playback volume of the at least two video clips includes:

Determine the difference in playback volume of any two adjacent video clips among the at least two video clips;

The second video decision result is determined according to the correlation between the playback amount difference and the difference threshold.

The difference threshold can be set according to the actual application, for example, the difference threshold is 500 or 1000, etc.

If at least two video segments include video segment 1, video segment 2, and video segment 3, any two adjacent video segments may be understood as video segment 1 and video segment 2, or video segment 2 and video segment 3.

Then after determining any two adjacent video clips, calculate the difference in playback volume of any two adjacent video clips; for example, any two adjacent video clips include video clip 1 and video clip 2, where the playback volume of video clip 1 The playback volume of video clip 2 is 100, and the playback volume of video clip 2 is 1100, then the playback volume difference between video clip 1 and video clip 2 is 1000; finally, according to the correlation between the playback volume difference and the preset difference threshold, Quickly determine the second video decision result to ensure the timeliness of video transcoding.

During specific implementation, determining the second video decision result based on the correlation between the playback amount difference and the difference threshold includes:

If the playback amount difference is greater than or equal to the difference threshold, determine that the second video decision result is video transcoding; or

If the playback amount difference is less than the difference threshold, it is determined that the second video decision result is that the video is not transcoded.

Following the above example, taking the play volume difference as 1000 and the difference threshold as 1000, it can be determined that the play volume difference of 1000 is equal to the difference threshold of 1000, then it can be determined that the second video decision result is video transcoding; and if When the difference in playback volume is 900 and the difference threshold is 1000, it can be determined that the playback volume difference 900 is less than the difference threshold 1000, and it can be determined that the second video decision result is that the video is not transcoded.

In practical applications, you can start playing the target video, that is, split a video clip every three minutes and obtain the current playback volume of the video clip. After dividing the second video clip and obtaining the current playback volume of the video clip, If it is determined that the growth rate of the second video clip is larger than that of the first video clip, the target video can be video transcoded; and by analogy, real-time playback and timely video transcoding can be achieved. code judgment.

In the video processing method provided by the embodiment of the present application, the second video can be quickly and accurately determined based on the correlation between the playback volume difference of any two adjacent video clips and the difference threshold during the playback of the target video. Decision result; that is, when the playback volume of the latter video clip has increased significantly compared with the playback volume of the previous video clip, it can be determined that the probability of the target video being played in large quantities is high. At this time, you can Video transcode the target video.

Alternatively, the playback amount of a continuous preset number of video clips can also be compared with the calculated popularity threshold, and the second video decision result can be quickly determined based on the relationship between the playback amount and the popularity threshold. The specific implementation method is as follows:

Determining the amount of playback of each of the at least two video clips;

Determine the popularity threshold of the at least two video clips based on the playback volume of each video clip;

The second video decision result is determined according to the play amount of each video segment in the at least two video segments and the popularity threshold.

The at least two video clips in this embodiment of the present application can be understood as all video clips divided according to the preset division rules within the target playback time period.

Then after obtaining the playback volume of each video clip, the popularity threshold for the target playback time period can be calculated based on the playback volume of all video clips; specifically, the calculation method of the popularity threshold is the same as the calculation method of the above characteristic threshold. are the same and will not be repeated here. For example, obtain the minimum value, average value, etc. of the playback volume of the video clip within the target playback time period, and perform calculations in the above manner to obtain the final popularity threshold.

After determining the popularity thresholds of at least two video clips based on the playback volume of each video clip, the second video decision result can be quickly determined based on the playback volume of the at least two video clips and the popularity threshold. The specific implementation method is as follows:

Determining the second video decision result based on the play amount of each video segment in the at least two video segments and the popularity threshold includes:

Obtain a preset number of consecutive video segments from the at least two video segments;

Determine the playback amount of each video segment in the continuous video segments;

The second video decision result is determined based on the play amount of each video segment in the continuous video segments and the correlation between the popularity threshold.

Among them, the preset number can be set according to the actual application, for example, the preset number can be set to 2, 3 or 4, etc.

Taking the preset number as 2 as an example, during specific implementation, two consecutive video clips are obtained from at least two video clips, and the playback amount of each video clip in the two consecutive video clips is determined; and then based on these two The correlation between the playback volume of each video clip in the consecutive video clips and the popularity threshold determines the second video decision-making result.

Still using the above example, two consecutive video clips include video clip 1 and video clip 2. Among them, the playback volume of video clip 1 is 100, the playback volume of video clip 2 is 1100, and the popularity threshold is 800. Then the follow-up can be based on The playback volume of video clip 1 is 100, and the playback volume of video clip 2 is 1100. They are respectively associated with the popularity threshold of 800, and the second video decision result is quickly obtained.

Among them, the specific implementation method of determining the second video decision result according to the play volume of each video segment in the consecutive video segments and the popularity threshold is as follows:

Determining the second video decision result based on the play amount of each video segment in the continuous video segments and the correlation between the popularity thresholds includes:

When the playback volume of each video segment in the continuous video segments is greater than or equal to the popularity threshold, determine that the second video decision result is video transcoding; or

When the play amount of any one of the continuous video segments is less than the popularity threshold, it is determined that the second video decision result is that the video is not transcoded.

Still using the above example, if the playback volume of video clip 1 is 1000, the playback volume of video clip 2 is 1500, and the popularity threshold is 800;

Then, by comparing the playback volume of each video clip in the continuous video clips with the popularity threshold, it can be determined that the playback volume of video clip 1 in the continuous video clips is 1000, and the playback volume of video clip 2 is 1500, both of which are greater than or equal to the popularity threshold. 800, at this time, it can be determined that the second video decision result is video transcoding. And if any one or both of video clip 1 and video clip 2 are less than the popularity threshold 800, it can be determined that the second video decision result is that the video is not transcoded.

Step 208: If the second video decision result satisfies the decision condition, perform video processing on the target video.

During specific implementation, the specific application scenarios of the video processing method provided by the embodiments of the present application are different, and the processing content of the video processing of the target video is also different; for example, the video processing method provided by the embodiments of the present application is applied to video transcoding. In the case of the scene, when the second video decision result satisfies the decision condition, video processing is performed on the target video, including:

When the second video decision result is video transcoding, it is determined that the second video decision result satisfies the decision condition, and the target video is transcoded.

Then when the decision condition is a video transcoding condition, after the second video decision result is determined through the above method, if the second video decision result is video transcoding, it can be determined that the second video decision result satisfies the video transcoding condition, At this point, the target video can be video transcoded.

The video processing method provided by the embodiment of the present application uses the object characteristics of the target object corresponding to the target video and a pre-trained object characteristic decision model to make a video transcoding decision in advance before the target video is played; and before the video is played, When the transcoding decision is that the video is not transcoded, after the target video is played, the video transcoding decision is made again through the preset video processing strategy to solve the waste of resources caused by transcoding all the target videos, and according to the above strategy When it is determined that the video needs to be transcoded, the target video should be transcoded effectively to avoid insufficient encoding and achieve accurate transcoding of the target video.

If the second video decision result does not meet the decision conditions, in order to maximize the difference between bandwidth revenue, computing power, and storage costs under limited resources, after the target video is played for a period of time, the video will be played again based on Based on the playback status of the target video that has been played, determine again whether the target video has transcoding value and whether video transcoding is required. The specific implementation method is as follows:

After determining the second video decision result according to the preset video processing strategy, the method further includes:

When the target playback time period ends and the second video decision result does not meet the decision condition, obtain the video playback characteristics of the target video within the preset playback time;

Input the video playback features into the playback feature decision model to obtain the third video decision result;

If the third video decision result satisfies the decision condition, perform video processing on the target video.

Among them, the preset playback time can be set according to time, and the preset playback time is greater than or equal to the target playback time period. For example, the target playback time period is 48 hours, and the preset playback time can be 50 hours or 60 hours. And the playback feature decision-making model is a pre-trained machine learning model, including but not limited to the XGBoost model. For the specific training process, please refer to the detailed introduction to the playback feature decision-making model training below.

Taking the target playback time period as 48 hours and the preset playback time as 50 hours as an example, when the target playback time period ends and the second video decision result does not meet the decision-making condition, the preset playback time is obtained. The video playback characteristics of the target video within the time period can be understood as, within 48 hours of playback of the target video, if the second video decision result obtained through any of the above preset video processing strategies still does not meet the decision conditions, Continue to play the target video, and when the target video plays for 50 hours, obtain the video playback characteristics of the target video during the 50 hours of playback.

In order to ensure the accuracy of video playback features and the rapid identification of subsequent playback feature decision-making models; the video playback features are obtained through data processing based on the obtained video playback attribute information of the target video within the preset playback time period. The specific implementation method is as follows:

The obtaining the video playback characteristics of the target video within the preset playback time period includes:

Obtain video playback attribute information of the target video within the preset playback time period;

Perform data processing on the video playback attribute information to obtain the video playback characteristics of the target video.

Among them, the video playback attribute information includes but is not limited to the number of views, reposts, likes, etc. after the target video is broadcast.

In practical applications, the specific processing method of performing data processing on the video playback attribute information to obtain the video playback characteristics of the target video is the same as the above-mentioned specific processing method of performing data processing on the object attribute information to obtain the object characteristics of the target object, and will not be used here. Again.

Specifically, after obtaining the video playback characteristics of the target video, the video playback characteristics can be input into the playback characteristics decision model to obtain the third video decision result; when the third video decision result satisfies the decision conditions, the target video can be Perform video processing; such as transcoding videos.

In practical applications, due to the preset video processing strategy and playback feature decision model, the video decision results obtained are after the target video is played, so if the video decision results obtained are video transcoding, in order to avoid causing video transcoding Code omission, when the target video that has not been played is transcoded based on the video decision result, the target video that has been played before the video transcoding is determined will also be supplemented and transcoded.

In the embodiment of this specification, when neither the first video decision result obtained through the above-mentioned object feature decision model nor the second video decision result obtained according to the preset video processing strategy satisfies the video transcoding conditions, in order to When resources are limited, the difference between bandwidth revenue, computing power, and storage costs is maximized. After the target video is played for a period of time, it will be judged again whether the target video has transcoding based on the playback status of the target video that has been played. Value, whether video transcoding is required to ensure accurate and timely transcoding of the target video.

Before using the playback feature decision-making model, the playback feature decision-making model will be pre-trained to improve the accuracy and effectiveness of the results of the playback feature decision-making model. The specific training method of the playback feature decision model is as follows:

The training steps of the playback feature decision model are as follows:

Obtain sample videos and determine the video playback attribute information corresponding to each sample video;

Determine the training sample according to the video playback attribute information corresponding to the sample video;

Determine the sample label corresponding to the training sample according to the video playback amount in the video playback attribute information;

The playback feature decision model is trained according to the training samples and the sample labels.

For a detailed introduction to the sample videos and the video playback attribute information corresponding to each sample video, please refer to the above embodiment.

Moreover, during the training of the playback feature decision model, in order to improve the training effect of the playback feature decision model, the video playback attribute information will also be data processed to obtain the video playback features of standard sample videos, and the playback feature decision model will be trained. The specific implementation method is as follows:

Determining the training sample based on the video playback attribute information corresponding to the sample video includes:

Perform data processing on the video playback attribute information corresponding to the sample video to obtain the video playback characteristics of the sample video;

The video playback characteristics of the sample video are determined as training samples.

During the specific implementation, for the specific training process of the playback feature decision model and the data processing process of the video playback attribute information corresponding to the sample video, please refer to the specific implementation process of the object feature decision model in the above embodiment, which will not be discussed here. Repeat.

The video processing method provided by the embodiment of the present application uses the object characteristics of the target object corresponding to the target video and a pre-trained object characteristic decision model to make a video transcoding decision in advance before the target video is played; and before the video is played, When the transcoding decision is that the video is not transcoded, after the target video is played, the video transcoding decision can be determined again through the preset video processing strategy within the target playback time period. If the video transcoding decision is that the video is not transcoded, after the target video has been played for a period of time, the video transcoding decision will be further determined based on the playback situation of the played video and the playback feature decision model; to solve the problem The waste of resources caused by transcoding all target videos, and when the video is determined to be transcoded according to the above strategy, the target video is effectively transcoded to avoid insufficient encoding and achieve accurate transcoding of the target video.

The video processing method will be further described below with reference to Figure 3, taking the application of the video processing method provided by this application in a video transcoding scenario as an example. Among them, Figure 3 shows a processing flow chart of a video processing method applied to a video transcoding scenario provided by an embodiment of the present application, which specifically includes the following steps:

Step 302: Obtain the video to be transcoded.

Step 304: Determine the uploader of the video to be transcoded, and obtain the uploader's historical video playback volume.

Step 306: Based on the uploader's historical video playback volume, determine the proportion of the uploader's video playback volume exceeding 10,000.

Step 308: Determine whether the proportion of the uploader's video playback volume exceeding 10,000 is greater than or equal to the preset proportion threshold. If yes, perform step 310. If not, perform step 312.

The preset proportion threshold is the same as the characteristic threshold in the above embodiment, and will not be described again here.

Step 310: Video transcoding.

Step 312: Obtain the object characteristics of the uploader, input the object characteristics of the uploader into the first machine learning model, and obtain the first video transcoding result of the video to be transcoded.

Step 314: Determine whether the first video transcoding result meets the transcoding conditions. If yes, execute step 310. If not, execute step 316.

The method of obtaining the object characteristics of the uploader is the same as the method of obtaining the object characteristics of the target object in the above embodiment; the first machine learning model can be understood as the object characteristics decision model in the above embodiment.

Step 316: Play the video to be transcoded, and determine the second video transcoding result according to the preset video transcoding strategy within the target playback time period.

The preset video transcoding strategy can be understood as the preset video processing strategy in the above embodiment.

Step 318: Determine whether the second video transcoding result meets the transcoding conditions. If yes, execute step 320. If not, execute step 322.

Step 320: Transcode the video and supplement the transcoding.

Specifically, when the second video transcoding result meets the transcoding conditions, that is, when video transcoding is required, the unplayed video to be transcoded is transcoded, and the previously played video to be transcoded is supplemented with transcoding. .

Step 322: During the preset playback time period of the video to be transcoded, obtain the video playback characteristics of the video to be transcoded within the preset playback time period, and input the video playback characteristics into the second machine learning model to obtain the third Video transcoding results.

The video playback features of the video to be transcoded are obtained in the same manner as the video playback features of the target video in the above embodiment; the second machine learning model can be understood as the playback feature decision model in the above embodiment.

Step 324: Determine whether the third video transcoding result meets the transcoding conditions. If yes, execute step 320. If not, end.

The video processing method provided by the embodiments of this application uses machine learning models and statistical methods to analyze video-related historical data, accurately determine whether the video has transcoding value, realize effective transcoding of the video, and achieve bandwidth benefits under limited resources. Maximize the difference with computing power and storage costs. Specifically, multiple data sources are used to conduct data analysis through machine learning models and statistical methods, and a transcoding decision-making strategy and transcoding decision-making model (the first machine learning model) are designed to make video conversions in advance before the video is opened. Coding decision-making: after the video is opened (played), another transcoding decision-making strategy and a transcoding decision-making model are used to complement the video transcoding in a timely manner to achieve timely and effective video transcoding, and to avoid unnecessary video transcoding and waste of resources. At the same time, videos with transcoding value should be identified as early as possible to ultimately maximize the difference between revenue and cost. That is to say, the video processing method provided by this application can conduct data analysis through machine learning models and statistical methods, design multiple transcoding decision strategies and transcoding decision models, and make video transcoding decisions in advance before the video is opened. Make up decisions in a timely manner to achieve timely and effective video transcoding.

Corresponding to the above method embodiments, this application also provides an embodiment of a video processing device. Figure 4 shows a schematic structural diagram of a video processing device provided by an embodiment of this application. As shown in Figure 4, the device includes:

The first result obtaining module 402 is configured to input the object characteristics of the target object corresponding to the target video into the object characteristic decision model to obtain the first video decision result when the target video is not played;

The video playback module 404 is configured to play the target video when the first video decision result does not meet the decision conditions;

The second result obtaining module 406 is configured to determine the second video decision result according to the preset video processing strategy within the target playback time period;

The video processing module 408 is configured to perform video processing on the target video if the second video decision result satisfies the decision condition.

Optionally, the device also includes:

The third result acquisition module is configured as:

When the target playback time period ends and the second video decision result does not meet the decision-making condition, obtain the video playback characteristics of the target video within the preset playback time period;

Optionally, the third result obtaining module is further configured as:

Optionally, the second result obtaining module 406 is further configured as:

Determining the amount of playback of each of the at least two video clips;

Optionally, the second result obtaining module 406 is further configured as:

Optionally, the video processing module 408 is further configured to:

Optionally, the device also includes:

The first model training module is configured to: train the object feature decision-making model;

The training steps of the object feature decision-making model are as follows:

Optionally, the first model training module is further configured to:

Optionally, the device also includes:

The feature threshold acquisition module is configured as:

Optionally, the device also includes:

The fourth result acquisition module is configured as:

Optionally, the first result obtaining module 402 is further configured as:

Optionally, the device also includes:

The second model training module is configured to: train the playback feature decision-making model;

The training steps of the playback feature decision model are as follows:

Optionally, the second model training module is further configured as:

The video processing device provided by the embodiment of the present application makes a video transcoding decision in advance based on the object characteristics of the target object corresponding to the target video and a pre-trained object characteristic decision model before the target video is played; and when the video When the transcoding decision is that the video is not transcoded, after the target video is played, the video transcoding decision is made again through the preset video processing strategy to solve the waste of resources caused by transcoding all the target videos, and according to the above strategy When it is determined that the video needs to be transcoded, the target video should be transcoded effectively to avoid insufficient encoding and achieve accurate transcoding of the target video.

The above is a schematic solution of a video processing device in this embodiment. It should be noted that the technical solution of the video processing device and the technical solution of the above-mentioned video processing method belong to the same concept. For details that are not described in detail in the technical solution of the video processing device, please refer to the description of the technical solution of the above video processing method. .

Figure 5 shows a structural block diagram of a computing device 500 provided according to an embodiment of this specification. Components of the computing device 500 include, but are not limited to, memory 510 and processor 520 . The processor 520 is connected to the memory 510 through a bus 530, and the database 550 is used to save data.

Computing device 500 also includes an access device 540 that enables computing device 500 to communicate via one or more networks 560 . Examples of these networks include the Public Switched Telephone Network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communications networks such as the Internet. Access device 540 may include one or more of any type of network interface (eg, a network interface card (NIC)), wired or wireless, such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, Global Interconnection for Microwave Access ( Wi-MAX) interface, Ethernet interface, Universal Serial Bus (USB) interface, cellular network interface, Bluetooth interface, Near Field Communication (NFC) interface, etc.

In one embodiment of this specification, the above-mentioned components of the computing device 500 and other components not shown in FIG. 5 may also be connected to each other, such as through a bus. It should be understood that the structural block diagram of the computing device shown in FIG. 5 is for illustrative purposes only and does not limit the scope of this description. Those skilled in the art can add or replace other components as needed.

Computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet computer, personal digital assistant, laptop computer, notebook computer, netbook, etc.), a mobile telephone (e.g., smartphone ), a wearable computing device (e.g., smart watch, smart glasses, etc.) or other type of mobile device, or a stationary computing device such as a desktop computer or PC. Computing device 500 may also be a mobile or stationary server.

When the processor 520 executes the instructions, the steps of the video processing method are implemented.

The above is a schematic solution of a computing device in this embodiment. It should be noted that the technical solution of the computing device and the technical solution of the above-mentioned video processing method belong to the same concept. For details that are not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above video processing method.

An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions. When the instructions are executed by a processor, the steps of the video processing method as described above are implemented.

The above is a schematic solution of a computer-readable storage medium in this embodiment. It should be noted that the technical solution of the storage medium and the technical solution of the above-mentioned video processing method belong to the same concept. For details that are not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above video processing method.

The above has described specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desired results. Additionally, the processes depicted in the figures do not necessarily require the specific order shown, or sequential order, to achieve desirable results. Multitasking and parallel processing are also possible or may be advantageous in certain implementations.

The computer instructions include computer program code, which may be in the form of source code, object code, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , random access memory (RAM, RandomAccess Memory), electrical carrier signals, telecommunications signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or deleted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium Excludes electrical carrier signals and telecommunications signals.

It should be noted that for the convenience of description, the foregoing method embodiments are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence. Because according to this application, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily necessary for this application.

In the above embodiments, each embodiment is described with its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

The preferred embodiments of the present application disclosed above are only used to help explain the present application. Alternative embodiments are not described in all details, nor are the inventions limited to the specific embodiments described. Obviously, many modifications and variations are possible in light of the teachings of this application. This application selects and specifically describes these embodiments in order to better explain the principles and practical applications of this application, so that those skilled in the art can better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

A video processing method including:

When the target video is not played, input the object characteristics of the target object corresponding to the target video into the object characteristics decision model to obtain the first video decision result;

When the first video decision result does not meet the decision conditions, play the target video;

Within the target playback time period, determine the second video decision result according to the preset video processing strategy;

If the second video decision result satisfies the decision condition, perform video processing on the target video.
The video processing method according to claim 1, after determining the second video decision result according to the preset video processing strategy, further comprising:

When the target playback time period ends and the second video decision result does not meet the decision-making condition, obtain the video playback characteristics of the target video within the preset playback time period;

Input the video playback features into the playback feature decision model to obtain the third video decision result;

If the third video decision result satisfies the decision condition, perform video processing on the target video.
According to the video processing method of claim 2, said obtaining the video playback characteristics of the target video within the preset playback time period includes:

Obtain video playback attribute information of the target video within the preset playback time period;

Perform data processing on the video playback attribute information to obtain the video playback characteristics of the target video.
The video processing method according to claim 1, wherein determining the second video decision result according to a preset video processing strategy includes:

At least two video clips are obtained according to the preset division rules, and the second video decision result is determined based on the playback volume of the at least two video clips.
The video processing method according to claim 4, wherein determining the second video decision result based on the playback volume of the at least two video clips includes:

Determine the difference in playback volume of any two adjacent video clips among the at least two video clips;

The second video decision result is determined according to the correlation between the playback amount difference and the difference threshold.
The video processing method according to claim 5, wherein determining the second video decision result based on the correlation between the playback volume difference and the difference threshold includes:

If the playback amount difference is greater than or equal to the difference threshold, determine that the second video decision result is video transcoding; or

If the playback amount difference is less than the difference threshold, it is determined that the second video decision result is that the video is not transcoded.
The video processing method according to claim 4, wherein determining the second video decision result based on the playback volume of the at least two video clips includes:

Determining the amount of playback of each of the at least two video clips;

Determine the popularity threshold of the at least two video clips based on the playback volume of each video clip;

The second video decision result is determined according to the play amount of each video segment in the at least two video segments and the popularity threshold.
The video processing method according to claim 7, wherein determining the second video decision result based on the play amount of each video segment in the at least two video segments and the popularity threshold includes:

Obtain a preset number of consecutive video segments from the at least two video segments;

Determine the playback amount of each video segment in the continuous video segments;

The second video decision result is determined based on the play amount of each video segment in the continuous video segments and the correlation between the popularity threshold.
The video processing method according to claim 8, wherein determining the second video decision result based on the play amount of each video segment in the continuous video segments and the correlation between the popularity thresholds includes:

When the playback volume of each video segment in the continuous video segments is greater than or equal to the popularity threshold, determine that the second video decision result is video transcoding; or

When the play amount of any one of the continuous video segments is less than the popularity threshold, it is determined that the second video decision result is that the video is not transcoded.
The video processing method according to claim 6 or 9, said performing video processing on the target video when the second video decision result satisfies the decision condition, including:

When the second video decision result is video transcoding, it is determined that the second video decision result satisfies the decision condition, and the target video is transcoded.
According to the video processing method of claim 1, the training steps of the object feature decision model are as follows:

Obtain sample videos, determine the sample objects corresponding to each sample video, and the video playback volume;

Determine the training sample according to the object attribute information of the sample object;

Determine the sample label corresponding to the training sample according to the video playback volume;

The object feature decision model is trained according to the training samples and the sample labels.
The video processing method according to claim 11, wherein determining the training sample according to the object attribute information of the sample object includes:

Perform data processing on the object attribute information of the sample object to obtain the object characteristics of the sample object;

The object characteristics of the sample object are determined as training samples.
The video processing method according to claim 12, after determining the sample label corresponding to the training sample according to the video playback amount, further comprising:

Determine the positive sample video in the sample video according to the sample label;

Determine the positive sample object corresponding to the positive sample video, and determine the target characteristics according to the historical video data of the positive sample object;

The corresponding feature threshold is determined according to the target feature.
The video processing method according to claim 13, before inputting the object characteristics of the target object into the object characteristic decision model and obtaining the first video decision result, it further includes:

Determine the target feature of the target object corresponding to the target video, and the characteristic value of the target feature;

Obtain a fourth video decision result according to the correlation between the characteristic value of the target characteristic and the characteristic threshold;

When the fourth video decision result satisfies the decision condition, video processing is performed on the target video according to the fourth video decision result.
The video processing method according to claim 1, wherein the object characteristics of the target object corresponding to the target video are input into the object characteristics decision model to obtain the first video decision result, including:

Obtain object attribute information of the target object corresponding to the target video;

Perform data processing on the object attribute information to obtain the object characteristics of the target object;

The object characteristics of the target object are input into the object characteristics decision model to obtain the first video decision result.
According to the video processing method of claim 2, the training steps of the playback feature decision model are as follows:

Obtain sample videos and determine the video playback attribute information corresponding to each sample video;

Determine the training sample according to the video playback attribute information corresponding to the sample video;

Determine the sample label corresponding to the training sample according to the video playback amount in the video playback attribute information;

The playback feature decision model is trained according to the training samples and the sample labels.
The video processing method according to claim 16, wherein determining the training sample based on the video playback attribute information corresponding to the sample video includes:

Perform data processing on the video playback attribute information corresponding to the sample video to obtain the video playback characteristics of the sample video;

The video playback characteristics of the sample video are determined as training samples.
A video processing device including:

The first result obtaining module is configured to input the object characteristics of the target object corresponding to the target video into the object characteristic decision model to obtain the first video decision result when the target video is not played;

A video playback module configured to play the target video when the first video decision result does not meet the decision condition;

The second result obtaining module is configured to determine the second video decision result according to the preset video processing strategy within the target playback time period;

A video processing module configured to perform video processing on the target video if the second video decision result satisfies the decision condition.
A computing device, including a memory, a processor, and computer instructions stored in the memory and executable on the processor. When the processor executes the instructions, the video processing method of any one of claims 1-17 is implemented. step.
A computer-readable storage medium stores computer instructions that, when executed by a processor, implement the steps of the video processing method described in any one of claims 1-17.