Welcome to our MODERATION page
Our AI model achieves over 90% success in content moderation by using advanced machine learning algorithms trained on vast datasets. These algorithms can accurately detect and classify inappropriate content based on predefined rules, ensuring high accuracy and consistency in moderation.
All moderation options focus on rejecting content with sexual activity, sexual displays, erotica, and sex toys. They also strictly reject offensive and hate content, including symbols associated with hate groups, and content related to recreational or medical drugs and many other options.
For Clients who have purschased the option, our default moderation model is setup to disallow poor quality images and videos, offensive or gore content and moderate QRcodes detected in images. These settings are pemanent and cannot be changed.
But paying clients can select multiple options to fine tune our model to their needs, Nudity (soft /hard), alcohol, smoking, money, gambling, advertising, presence of children or minors are amongst the possible options. This allows a very adaptative model that will fit your policies and your event
Moderating guest-generated images for public or live display is crucial to protect viewers, ensure
- legal compliance,
- maintain brand reputation,
- and prevent abuse,
guaranteeing a safe and respectful live viewing experience for everyone.
The permanent rules of our moderation model are designed to ensure a high standard of content quality and safety. Firstly, the model detects and rejects poor quality images, such as those that are blurred or of low resolution. It also identifies and rejects images that are not natural photos. Offensive content, including gestures, symbols, and flags, is strictly prohibited and will be rejected. The model also screens for and rejects horrific imagery, which includes blood, guts, self-harm, wounds, and human skulls. QR codes detected within images are moderated to prevent inappropriate content from being shared through this medium. Lastly, any images identified as part of romance scams are automatically detected and rejected. These permanent rules cannot be modified, ensuring consistent and reliable moderation.
All moderated images are sent to your backend in a blurred format and are never visible in the public gallery or live slider.
This setup allows you to check the quality of the moderation algorithm, as you can unblur the images for review. Each passed or rejected image in your backend (client panel) is annotated with the results of the selected moderation tests.
If a rule has been applied, you are systematically informed of the specific rule that triggered the moderation.
The moderation model offers various optional checks that clients can customize according to their needs.
It can detect and reject adult content, including nudity, sexual activity, and suggestive scenes. However, it can bypass and allow soft adult content, such as revealing cleavage and miniskirts. For artworks, the model can bypass adult content tests and allow adult content if it is part of the artwork.
Additionally, the model can reject images containing weapons, including guns, rifles, substance abuse, and threatening knives. It can also identify and reject images containing advertisements. Images where people are under the influence of substances can be detected and rejected as well.
The model is capable of detecting casino scenes, gambling, and slot machines, and it can reject images with artificial text embedded. Smoking-related content, including cigarettes, cigars, and pipes, can be detected and rejected.
For alcohol-related content, the model can detect and reject images containing alcoholic beverages but can bypass this check if only glasses are visible and their content is not obviously alcohol. It can also detect and reject images displaying money and banknotes.
It can identify and reject images containing celebrities.
Furthermore, the model can be asked to reject images if a child under 12 or a minor under 18 is present. These optional checks allow for a highly adaptable moderation model that can fit various policies and event requirements.
For paying clients, our model offers extensive customization options. Clients can select multiple options to fine-tune our model to their needs. These customizable options allow our AI model to adapt to various client policies and event requirements, ensuring a tailored and effective content moderation solution.
Copyright © ZAAGS.com 2023