tencent cloud

Media Processing Service

Release Notes and Announcements
Release Notes
Announcements
Product Introduction
Overview
Features
Strengths
Use Cases
Purchase Guide
Billing Overview
Purchase Instructions
Top Up and Purchase
Overdue Payments
Refund
Getting Started
Console Guide
Overview
Creating Tasks
Tasks
Orchestrations
Templates
Resource Packs
Video Evaluation
AIGC Content Generation
Terminal SDK
Subtitle Editing Tool
Usage Statistics
Cloud Access Management Sample
Integration Tutorials
Audio/Video Transcoding Integration
Audio/Video Enhancement Integration
Audio Separation Integration
Integration of Digital and Visible Watermarks
Media AI Integration Tutorial
Media Quality Inspection Integration
Terminal SDK integration
MPS Live Stream Recording integration
DRM integration
Other tutorials
Application Scenario and Practical Tutorial
Image Quality Improvement Scenario
Audio/Video Cost Optimization Scenario
Short Drama Translation Scenario
AI-Generated Content Scenario
Online Education Scenarios
API Documentation
History
Introduction
API Category
Making API Requests
Processing Task Initiation APIs
Task Management APIs
Transcoding and Enhancement Template APIs
Watermark Template APIs
Screenshot Template APIs
Media AI Template APIs
Media AI—Hotword Lexicon APIs
Media AI—Sample Management APIs
Media Quality Inspection Template APIs
Live Streaming Recording Template APIs
Orchestration Management APIs
Data Statistics APIs
StreamLink—Security Group Management APIs
Other APIs
Image Processing Template APIs
AI Generation APIs
Other APIs
Data Types
Error Codes
Other Documents
WebSocket Protocol for Recognition
FAQs
Basics
Account Authorization
Task Configuration
Task Initiation
Task Result Viewing
Related Agreement
Service Level Agreement
Privacy Policy
Data Processing And Security Agreement
Contact Us
Glossary
DocumentationMedia Processing Service

Intelligent Video Recognition

Focus Mode
Font Size
Last updated: 2022-01-18 16:17:11
MPS leverages AI technologies to recognize video content. The result of an intelligent video recognition task includes a recognition score, suggestion, and suspicious video segments. You can decide whether to expose a video based on the suggestion.

Result

MPS can intelligently recognize video images, speech (ASR), and optical characters (OCR).
Object
Operation
Description
Video images (People and objects)
Pornographic content
Checks for pornographic content in video images, including:`vulgar`: vulgarity`intimacy`: intimacy`sexy`: sexiness
Politically sensitive content
Checks for politically sensitive content in video images, including:`bloody`: bloodiness`explosion`: explosions and fires`violation_photo`: banned icons`guns`: weapons and guns
Speech (Speech to text)
Pornographic content
Checks for keywords for pornographic content in speech
Politically sensitive content
Checks for keywords for politically sensitive content in speech
Optical characters Image to text
Pornographic content
Checks for keywords for pornographic content in images
Politically sensitive content
Checks for keywords for politically sensitive content in images

Parameter description

Field Name
Type
Description
confidence
Float
Intelligent recognition score (0-100). The higher the score, the more suspicious the content.
suggestion
String
There are three types of intelligent recognition suggestions:
pass: The degree of suspicion is not high, and approval is recommended.
review: The degree of suspicion is high, and human review is recommended.
block: The degree of suspicion is very high, and blocking is recommended.
segments
Array
Suspicious video segments, which help you locate suspicious content in a video

Initiating Task

Directions

You can call an API to initiate an intelligent video recognition task or configure automatic triggering of the task upon video upload.
API: Call the ProcessMedia API, setting AiContentReviewTask to the ID of your intelligent video recognition template.
Automatic triggering upon upload: In the console, create a workflow with intelligent video recognition enabled and upload videos to the bucket bound to the workflow.

Creating template

MPS uses templates to represent combinations of intelligent video recognition parameters, which determine which of the following operations MPS performs.
Recognition of pornographic content in video images
Recognition of politically sensitive content in video images
Recognition of pornographic keywords in speech (ASR)
Recognition of politically sensitive keywords in speech (ASR)
Recognition of pornographic keywords in images (OCR)
Recognition of politically sensitive keywords in images (OCR)
MPS provides preset intelligent video recognition templates for common parameter combinations. You can also use a server API to create and manage custom templates.

Obtaining Result

After initiating an intelligent video recognition task, you can wait for the result notification asynchronously or query the result synchronously.
Below is an example of the result returned after query (fields with null values are omitted):
{
"TaskType":"WorkflowTask",
"Status":"FINISH",
"CreateTime":"2019-07-16T06:21:27Z",
"BeginProcessTime":"2019-07-16T06:21:28Z",
"FinishTime":"2019-07-16T06:21:46Z",
"WorkflowTask":{
"TaskId":"2356768367-WorkflowTask-2e1af2456351812be963e309cc133403t0",
"Status":"FINISH",
"InputInfo":{
"Type":"COS",
"CosInputInfo":{
"Bucket":"MyVideoBucket-235303****",
"Region":"ap-beijing",
"Object":"/input/AnimalWorld.mp4"
}
},
"MetaData":{
"AudioDuration":60,
"AudioStreamSet":[
{
"Bitrate":383854,
"Codec":"aac",
"SamplingRate":48000
}
],
"Bitrate":1021028,
"Container":"mov,mp4,m4a,3gp,3g2,mj2",
"Duration":60,
"Height":480,
"Rotate":0,
"Size":7700180,
"VideoDuration":60,
"VideoStreamSet":[
{
"Bitrate":637174,
"Codec":"h264",
"Fps":23,
"Height":480,
"Width":640
}
],
"Width":640
},
"MediaProcessResultSet":[

],
"AiContentReviewResultSet":[
{
"Type":"Porn",
"PornTask":{
"Status":"SUCCESS",
"ErrCode":0,
"Message":"",
"Input":{
"Definition":10
},
"Output":{
"Confidence":98,
"Suggestion":"block",
"Label":"sexy",
"SegmentSet":[
{
"StartTimeOffset":9.5,
"EndTimeOffset":14,
"Confidence":98,
"Suggestion":"block",
"Label":"sexy",
"Url":"http://xxx.vod2.myqcloud.com/xxx/xxx/xx1.jpg",
"PicUrlExpireTime":"2019-07-23T06:21:46Z"
},
{
"StartTimeOffset":16.5,
"EndTimeOffset":18,
"Confidence":80,
"Suggestion":"review",
"Label":"sexy",
"Url":"http://xxx.vod2.myqcloud.com/xxx/xxx/xx2.jpg",
"PicUrlExpireTime":"2019-07-23T06:21:46Z"
},
{
"StartTimeOffset":41,
"EndTimeOffset":49,
"Confidence":97,
"Suggestion":"block",
"Label":"sexy",
"Url":"http://xxx.vod2.myqcloud.com/xxx/xxx/xx3.jpg",
"PicUrlExpireTime":"2019-07-23T06:21:46Z"
}
]
}
}
},
{
"Type":"Terrorism",
"TerrorismTask":{
"Status":"SUCCESS",
"ErrCode":0,
"Message":"",
"Input":{
"Definition":10
},
"Output":{
"Confidence":0,
"Suggestion":"pass",
"SegmentSet":[

]
}
}
},
{
"Type":"Political",
"PoliticalTask":{
"Status":"SUCCESS",
"ErrCode":0,
"Message":"",
"Input":{
"Definition":10
},
"Output":{
"Confidence":0,
"Suggestion":"pass",
"SegmentSet":[

]
}
}
}
],
"AiAnalysisResultSet":[

],
"AiRecognitionResultSet":[

]
},
"TasksPriority":0,
"SessionId":"",
"SessionContext":"",
"RequestId":"xxx-xxx-xxx"
}
As shown above, there are three types of results under WorkflowTask.AiContentReviewResultSet: Porn, Terrorism, and Political.
For Porn, Output.Suggestion is block, which indicates a very high likelihood that the content is pornographic, and you are advised to block it. The confidence score is 98, and the label for the content is sexy.
Three suspicious video segments are identified for Porn, whose start and end times are specified by StartTimeOffset and EndTimeOffset.
According to the results for Terrorism and Political, no inappropriate content is detected in the video.

Help and Support

Was this page helpful?

Help us improve! Rate your documentation experience in 5 mins.

Feedback