tencent cloud

Feedback

Batch Image Moderation

Last updated: 2024-06-12 16:01:53

    Feature Overview

    The batch image moderation API adopts a sync or async POST request method. You can use this API to perform content moderation on multiple image files.
    Note:
    You can moderate image files stored in COS.
    You can moderate images at URLs of a third-party cloud storage vendor.
    To moderate encrypted data, please contact us.
    You can customize moderation policies based on different business scenarios as instructed in Setting Moderation Policy.

    Restrictions

    Supported image file size: < 32 MB. To moderate images larger than 5 MB in size, you need to use the large-image-detect parameter when calling the API.
    Supported image quantity: Up to 100 images per request.
    Supported image file resolution: a resolution of 256x256 or higher is recommended; otherwise, the recognition effect may be affected.
    Supported image file format: PNG, JPG/JPEG, BMP, GIF, WEBP, or HEIF/HEIC.
    Supported image URL transfer protocols: HTTP and HTTPS.
    Calling the API requires a signature. For more information, see Request Signature.

    SDK Recommendation

    COS SDK provides complete capabilities of demo, automatic integration, and signature calculation. You can easily and quickly call APIs through the SDK. For more information, see SDK Overview.

    Request

    Sample request

    POST /image/auditing HTTP/1.1
    Host: <BucketName-APPID>.ci.<Region>.myqcloud.com
    Date: <GMT Date>
    Authorization: <Auth String>
    Content-Length: <length>
    Content-Type: application/xml
    
    <body>
    Note:
    Authorization: Auth String (see Request Signature for more information)
    When this feature is used by a sub-account, relevant permissions must be granted as instructed in Authorization Granularity Details.

    Request headers

    This API only uses Common Request Headers.

    Request body

    This request requires the following request body:
    <Request>
    <Input>
    <DataId></DataId>
    <Url></Url>
    <Object></Object>
    <Content></Content>
    <Interval></Interval>
    <MaxFrames></MaxFrames>
    <LargeImageDetect></LargeImageDetect>
    <UserInfo>
    <TokenId></TokenId>
    </UserInfo>
    </Input>
    <Input>
    <DataId></DataId>
    <Url></Url>
    <Object></Object>
    <Content></Content>
    <Interval></Interval>
    <MaxFrames></MaxFrames>
    </Input>
    <Conf>
    <BizType>b81d45f94b91a683255e9a9506f45a11</BizType>
    <Async></Async>
    <Callback></Callback>
    <Freeze>
    <PornScore></PornScore>
    </Freeze>
    </Conf>
    </Request>
    The nodes are described as follows:
    Node Name (Keyword)
    Parent Node
    Description
    Type
    Required
    Request
    None
    Batch image moderation configuration.
    Container
    Yes
    Request has the following sub-nodes:
    Node Name (Keyword)
    Parent Node
    Description
    Type
    Required
    Input
    Request
    Content to be moderated. If there are multiple images, pass in multiple Input structures.
    Container Array
    Yes
    Conf
    Request
    Moderation rule configuration.
    Container
    Yes
    Input has the following sub-nodes, and you need to use only one of these sub-nodes:
    Node Name (Keyword)
    Parent Node
    Description
    Type
    Required
    Object
    Request.Input
    Name of the image file stored in the COS bucket; for example, if the file is image.jpg in the test directory, then the filename is test/image.jpg. You can select only one parameter among Content, Object, and Url at a time. If more than one parameter is passed in, only one parameter takes effect (in the priority of Content, Object, and Url).
    String
    No
    Url
    Request.Input
    Full URL of the image file, such as http://a-1250000.cos.ap-shanghai.myqcloud.com/image.jpg. You can select only one parameter among Content, Object, and Url at a time. If more than one parameter is passed in, only one parameter takes effect (in the priority of Content, Object, and Url).
    String
    No
    Content
    Request.Input
    Image file content, which needs to be Base64-encoded first. If multiple parameters among Content, Object, and Url are passed in, only one will take effect in the order of Content > Object > Url.
    String
    No
    Interval
    Request.Input
    Frame capturing frequency, which takes effect for GIF images only. The default value is 5, indicating to capture a frame every five frames starting from the first frame (included).
    Int
    No
    MaxFrames
    Request.Input
    The maximum number of frames to be captured, which takes effect for GIF images only. The default value is 5, indicating to capture only five frames of the GIF image for moderation. The parameter value must be greater than 0.
    Int
    No
    DataId
    Request.Input
    Image ID. This field will return the original content in the result, which can contain up to 512 bytes.
    String
    No
    LargeImageDetect
    Request.Input
    Whether to compress the image that exceeds the size limit before moderation. Valid values: 0 (no), 1 (yes). Default value: 0. Note: Images up to 32 MB in size can be compressed, and compression fees will be charged.
    Int
    No
    UserInfo
    Request.Input
    Business field.
    Container
    No
    UserInfo has the following sub-nodes:
    Node Name (Keyword)
    Description
    Type
    Required
    TokenId
    Account information, which can contain up to 128 bytes.
    String
    No
    Nickname
    Nickname information, which can contain up to 128 bytes.
    String
    No
    DeviceId
    Device information, which can contain up to 128 bytes.
    String
    No
    AppId
    Unique app ID, which can contain up to 128 bytes.
    String
    No
    Room
    Room ID information, which can contain up to 128 bytes.
    String
    No
    IP
    IP address information, which can contain up to 128 bytes.
    String
    No
    Type
    Business type, which can contain up to 128 bytes.
    String
    No
    ReceiveTokenId
    User account to receive messages, which can contain up to 128 bytes.
    String
    No
    Gender
    Gender information, which can contain up to 128 bytes.
    String
    No
    Level
    Level information, which can contain up to 128 bytes.
    String
    No
    Role
    Role information, which can contain up to 128 bytes.
    String
    No
    Conf has the following sub-nodes:
    Node Name (Keyword)
    Parent Node
    Description
    Type
    Required
    BizType
    Request.Conf
    Unique identifier of the moderation policy. You can configure the scenes you want to moderate on the moderation policy page in the console, such as pornographic, advertising, and illegal information. For configuration guidelines, see Setting Moderation Policy. You can get BizType in the console. If BizType is specified, the moderation request will perform moderation based on the scenes configured in the moderation policy. If BizType is not specified, the default moderation policy will be used automatically.
    String
    No
    Async
    Request.Conf
    Whether to moderate asynchronously. Valid values: 0 (returns the result synchronously); 1 (moderates asynchronously). Default value: 0.
    Integer
    No
    Callback
    Request.Conf
    The moderation result (in Detail mode) can be sent to your callback address in the form of a callback. This parameter takes effect in async moderation. Addresses starting with http:// or https:// are supported, such as http://www.callback.com.
    String
    No
    Freeze
    Request.Conf
    This field can be used to set automatic freezing for images based on the moderation score. It takes effect only if the image moderated in input is an object.
    Container
    No
    Freeze has the following sub-nodes:
    Node Name (Keyword)
    Parent Node
    Description
    Type
    Required
    PornScore
    Request.Conf.Freeze
    The threshold at or above which automatic freezing will be performed for the porn moderation result. Value range: [0,100]. If this field is left empty (default value), automatic freezing will not be performed.
    Integer
    No
    AdsScore
    Request.Conf.Freeze
    The threshold at or above which automatic freezing will be performed for the ad moderation result. Value range: [0,100]. If this field is left empty (default value), automatic freezing will not be performed.
    Integer
    No
    For freezing parameters in other moderation scenes, contact the customer service.
    Note:
    Moderation through Object is a private network operation and will not generate public network traffic.
    Moderation through Url will generate public network traffic with regard to the origin where the image resides.

    Response

    Response headers

    This API only returns Common Response Headers.

    Response body

    The response body returns application/xml data. The following contains all the nodes:
    <Response>
    <JobsDetail>
    <DataId></DataId>
    <Url></Url>
    <PornInfo>
    <HitFlag>1</HitFlag>
    <Score>100</Score>
    <Category>xxx</Category>
    <SubLabel>xxx</SubLabel>
    </PornInfo>
    </JobsDetail>
    <JobsDetail>
    <Code>InvalidArgument</Code>
    <Message>Param MaxFrames is illegal</Message>
    <DataId></DataId>
    </JobsDetail>
    </Response>
    The response body is as described below:
    Node Name (Keyword)
    Parent Node
    Description
    Response
    Container
    The specific response content returned by batch image moderation.
    Response has the following sub-nodes:
    Parameter
    Type
    Description
    JobsDetail
    Container Array
    Image moderation result.
    RequestId
    String
    ID automatically generated by the server for a request when the request is sent, which can be used to facilitate fault locating.
    The async moderation result JobsDetail has the following sub-nodes:
    Node Name (Keyword)
    Type
    Description
    Code
    String
    Error code, which will be returned only in case of a failure. For more information, see Error Codes.
    Message
    String
    Error message, which will be returned only in case of a failure.
    DataId
    String
    Image ID. The original content will be returned in the moderation result, which can contain up to 512 bytes.
    JobId
    String
    Moderation job ID.
    State
    String
    Status of the moderation job. Valid values: Submitted, Failed.
    Object
    String
    The name of the image file stored in the COS bucket, which will be returned if Object is selected during job creation.
    Url
    String
    The URL of the image file, which will be returned if Url is selected during job creation.
    The sync moderation result JobsDetail has the following sub-nodes:
    Parameter
    Type
    Description
    Code
    String
    Error code, which will be returned only in case of a failure. For more information, see Error Codes.
    Message
    String
    Error message, which will be returned only in case of a failure.
    DataId
    String
    Image ID. The original content will be returned in the moderation result, which can contain up to 512 bytes.
    JobId
    String
    Moderation job ID.
    State
    String
    Status of the moderation job. Valid values: Success, Failed.
    Object
    String
    The name of the image file stored in the COS bucket, which will be returned if Object is selected during job creation.
    Url
    String
    The URL of the image file, which will be returned if Url is selected during job creation.
    CompressionResult
    Integer
    Whether the image is compressed. Valid values: 0 (no), 1 (yes).
    Label
    String
    This field is used to return the maliciousness label with the highest priority in the detection result, which represents the moderation result suggested by the model. We recommend you handle different types of violations and suggestions based on your business needs. Return values: Normal, Porn, Ads.
    Result
    Integer
    This field indicates the moderation result. You can perform subsequent operations based on the result. We recommend you handle different results based on your business needs. Valid values: 0 (normal), 1 (sensitive), 2 (suspiciously sensitive, with human review recommended).
    Score
    Integer
    The confidence the moderation result hits the moderation scene. Value range: 0–100. The higher the value, the more likely the content hits the currently returned moderation scene. For example, Porn 99 means that the content is very likely to be pornographic.
    Category
    String
    Moderation type hit by the image.
    SubLabel
    String
    Second-level label hit by the image.
    Text
    String
    The specific text content recognized by OCR in the image, which will be returned if text content detection is enabled in the moderation policy.
    PornInfo
    Container
    The moderation result of the pornographic information moderation scene.
    AdsInfo
    Container
    The moderation result of the advertising information moderation scene.
    UserInfo
    Container
    UserInfo set in the request, which will be returned as-is.
    ListInfo
    Container
    Blocklist/Allowlist status of the account.
    ForbidState
    Integer
    If you set automatic freezing during request, this field indicates the status of the image. 0: not frozen; 1: frozen.
    The moderation information (such as PornInfo and AdsInfo) has the following sub-nodes:
    Parameter
    Type
    Description
    Code
    Integer
    Error code for individual moderation scenes. 0: succeeded; other values: failed. For more information, see Error Codes.
    Msg
    String
    The specific error message, which will be OK if the moderation result is normal.
    HitFlag
    Integer
    The moderation result returned for the moderation scene. Return values: 0 (normal); 1 (confirmed as a violation of the current scene); 2 (suspected as a violation of the current scene).
    Score
    Integer
    The confidence the moderation result hits the moderation scene. Value range: 0–100. The higher the value, the more likely the content hits the currently returned moderation scene. For example, Porn 99 means that the content is very likely to be pornographic.
    Label
    String
    The overall result label of the image, which may be SubLabel, a person name, etc.
    Category
    String
    This field is a subset of Label, indicating the specific moderation category hit; for example, Sexy represents the sexy category in the Porn label.
    SubLabel
    String
    Second-level label of the image.
    OcrResults
    Container Array
    This field represents the detailed OCR result, including the text coordinate information and text recognition result. It will be returned if there is non-compliant content.
    LibResults
    Container Array
    This field returns results based on recognition against the risk library. Note: This field will not be returned if no samples in the risk library are hit.
    LibResults has the following sub-nodes:
    Node Name (Keyword)
    Type
    Description
    ImageId
    String
    This field represents the hit image sample ID in the risk library.
    Score
    Integer
    This field returns the confidence under the current label. Value range: 0–100. The higher the value, the more likely the image hits a sample in the risk library. For example, Porn 99 means that the content is very likely to hit a pornographic sample in the library.
    OcrResults has the following sub-nodes:
    Node Name (Keyword)
    Type
    Description
    Text
    String
    Sensitive OCR text content
    Keywords
    String Array
    Sensitive words in the text content
    Location
    Container
    Location information of the text content in the image
    ObjectResults has the following sub-nodes:
    Node Name (Keyword)
    Description
    Type
    Name
    String
    Sensitive content name
    Location
    Container
    Location information of the content in the image
    Location has the following sub-nodes:
    Node Name (Keyword)
    Type
    Description
    X
    Float
    X-coordinate relative to the origin (top-left corner of the image)
    Y
    Float
    Y-coordinate relative to the origin (top-left corner of the image)
    Height
    Float
    Height of the sensitive area
    Width
    Float
    Width of the sensitive area
    Rotate
    Float
    Rotation angle of the sensitive area
    UserInfo has the following sub-nodes (same as the UserInfo in the request):
    Node Name (Keyword)
    Description
    Type
    Required
    TokenId
    Account information, which can contain up to 128 bytes.
    String
    No
    Nickname
    Nickname information, which can contain up to 128 bytes.
    String
    No
    DeviceId
    Device information, which can contain up to 128 bytes.
    String
    No
    AppId
    Unique app ID, which can contain up to 128 bytes.
    String
    No
    Room
    Room ID information, which can contain up to 128 bytes.
    String
    No
    IP
    IP address information, which can contain up to 128 bytes.
    String
    No
    Type
    Business type, which can contain up to 128 bytes.
    String
    No
    ReceiveTokenId
    User account to receive messages, which can contain up to 128 bytes.
    String
    No
    Gender
    Gender information, which can contain up to 128 bytes.
    String
    No
    Level
    Level information, which can contain up to 128 bytes.
    String
    No
    Role
    Role information, which can contain up to 128 bytes.
    String
    No
    ListInfo has the following sub-nodes:
    Node Name (Keyword)
    Type
    Description
    ListResults
    Container Array
    Results in all hit lists.
    ListResults has the following sub-nodes:
    Node Name (Keyword)
    Description
    Type
    ListType
    Integer
    Type of the hit list. Valid values: 0 (allowlist), 1 (blocklist).
    ListName
    String
    Name of the hit list.
    Entity
    String
    Entity hit in the list.

    Use Cases

    Request

    POST /image/auditing HTTP/1.1
    Authorization: q-sign-algorithm=sha1&q-ak=AKIDZfbOAo7cllgPvF9cXFrJD0**********&q-sign-time=1497530202;1497610202&q-key-time=1497530202;1497610202&q-header-list=&q-url-param-list=&q-signature=28e9a4986df11bed0255e97ff90500557e0e****
    Host: examplebucket-1250000000.cos.ap-chengdu.myqcloud.com
    Content-Length: 166
    Content-Type: application/xml
    
    <Request>
    <Input>
    <Object>a.jpg</Object>
    </Input>
    <Input>
    <Url>http://exsample.com/b.jpg</Url>
    </Input>
    <Conf>
    <BizType>b81d45f94xxxxxxxxa9506f45a11</BizType>
    </Conf>
    </Request>

    Response

    HTTP/1.1 200 OK
    Content-Type: application/xml
    Content-Length: 645
    Date: Tue, 03 Apr 2018 09:06:16 GMT
    Status: 200 OK
    x-cos-request-id: NWFjMzQ0MDZfOTBmYTUwXzZkZV8z****
    
    <Response>
    <JobsDetail>
    <JobId>xxxxxxxxxx</JobId>
    <CompressionResult>0</CompressionResult>
    <Object>a.jpg</Object>
    <Label>Porn</Label>
    <Result>1</Result>
    <Score>100</Score>
    <SubLabel>xxx</SubLabel>
    <PornInfo>
    <Code>0</Code>
    <Msg>OK</Msg>
    <HitFlag>1</HitFlag>
    <Score>100</Score>
    <Label>xxx</Label>
    <SubLabel>xxx</SubLabel>
    </PornInfo>
    </JobsDetail>
    <JobsDetail>
    <JobId>yyyyyyyyyy</JobId>
    <CompressionResult>0</CompressionResult>
    <Url>http://exsample.com/b.jpg</Url>
    <Label>Porn</Label>
    <Result>1</Result>
    <Score>100</Score>
    <SubLabel>xxx</SubLabel>
    <PornInfo>
    <Code>0</Code>
    <Msg>OK</Msg>
    <HitFlag>1</HitFlag>
    <Score>100</Score>
    <Label>xxx</Label>
    <SubLabel>xxx</SubLabel>
    </PornInfo>
    </JobsDetail>
    <RequestId>xxxxxxxxxxx</RequestId>
    </Response>
    
    Contact Us

    Contact our sales team or business advisors to help your business.

    Technical Support

    Open a ticket if you're looking for further assistance. Our Ticket is 7x24 avaliable.

    7x24 Phone Support