Overview

BOS image audit service is the external image intelligent audit service provided by Baidu AI Cloud. Based on the intelligent content audit scheme of in-depth learning, it can accurately identify yellow, violence, terrorism, political sensitivity, WeChat business advertising, nausea and other contents in images and videos. For your convenience, BOS provides developers with Image Review API or CLI call, and also provides an interface-based image review console. Users can view the violation photos in the BOS review results on the console, and conduct manual reviews as needed.

Notification

• To avoid violation image propagation, BOS will temporarily set the permission of the violation image to “private permission”, that is, only the Bucket owner can see it.
• If the image is generated by cross-region replication and life cycle policies in the Bucket with the image audit policy configured, it will not be automatically audited.
• For the un-retrieved Object of archive type, the image audit policy will not take effect; the retrieved Object of archive type can be audited normally.

Set Audit Rules

1.Log in BOS Management Console to enter the overview page. 2.Click the Use Now button in the "Image Review" card at the bottom of the global overview page, and click the Enable Photo Review button in the pop-up "Enable Image Review" confirmation box to enter the BOS image review console. 3.Click the Settings button in the upper right corner to Set Review Rules.

• Bucket review: Select the target Bucket to review.
• Review type: Choose which types of violations need to be reviewed. Currently, it provides three types of violation review capabilities: pornography, violence/terrorism, and political sensitivity.

• Review limit: Set the maximum number of reviews conducted per day through the photo review console to avoid excessive costs caused by adding excessively review images.

  4.Click OK to save the review rule settings.

Use Review Service

The BOS image audit service will monitor and audit all the new images in the BUCKET you select, and clearly display the audit information to you in the way of dashboard .

• Today's photo data: The audited image quantity of image data today represents the total number of audited images in the Bucket where you opened the image audit today.
• Number of illegal photo volume: Represent the number of illegal photos as the result of the review.
• Number of illegal photos manually handled: Represent the number of photos that the administrator has manually processed based on the number of illegal images.

The audited violation images will be uniformly displayed in the lower panel of the dashboard . You can select to display the violation image you want to review manually by filtering Bucket dimension and time dimension.

When you manually review the violation images on demand, you can view the details of the selected images, directly delete or restore the original permissions.

API Usage

You can also access the image review service through Image Review API.