Image Moderation
This documentation outlines the various types of image moderation available in Lasso Moderation. Each type helps ensure that the content on your platform is safe, compliant, and adheres to community guidelines.
Optical Character Recognition (OCR)
OCR (Optical Character Recognition) is used for extracting text from images. This is particularly useful for identifying inappropriate or harmful text in images, such as hate speech, explicit language, or spam. OCR can recognize various fonts and text layouts, making it a robust tool for detecting textual content within an image. Example:
How it works:
The image is scanned for areas containing text and extracted from the image.
This text can then be checked against moderation policies for further action.
Label Detection
Label detection is used to classify and categorize images into predefined categories and subcategories. Lasso Moderation provides a range of default categories and subcategories that help identify inappropriate content:
Explicit Nudity
Explicit Nudity
Explicit Sexual Activity
Sex Toys
Exposed Female Nipples
Exposed Buttocks
Exposed Male Genitalia
Exposed Female Genitalia
Non-Explicit Nudity of Intimate Parts and Kissing
Non-Explicit Nudity
Obstructed Intimate Parts
Implied Nudity
Exposed Male Nipple
Kissing on the Lips
Swimwear or Underwear
Female Swimwear or Underwear
Male Swimwear or Underwear
Violence
Graphic Violence
Blood & Gore
Physical Violence
Weapon Violence
Weapons
Self-Harm
Visually Disturbing
Death and Emaciation
Crashes
Emaciated Bodies
Corpses
Air Crash
Explosions and Blasts
Rude Gestures
Rude Gestures
Middle Finger
Drugs & Tobacco
Products
Pills
Drug & Tobacco Paraphernalia & Use
Smoking
Alcohol
Alcohol
Alcohol Use
Drinking
Alcoholic Beverages
Gambling
Hate Symbols
Hate Symbols
Nazi Party
White Supremacy
Extremism
Custom Labels
Lasso Moderation allows users to create custom labels for detecting specific types of objects or scenes. This feature enables the detection of items like "animals" or "objects" based on your platform's needs. Custom labels are useful for identifying specific content not covered by the default labels.
Examples of custom labels:
Animals (e.g., cats, dogs)
Vehicles (e.g., cars, trucks)
Specific objects (e.g., logos, tools)
Face Detection
The face detection feature identifies faces within images and provides additional information such as:
Age estimation
Gender prediction
Additionally, Lasso Moderation can optionally offer face similarity detection, which helps recognize images with specific people in it.
Celebrity Detection
This feature detects and identifies well-known celebrities in images. It uses a comprehensive database of public figures to provide accurate recognition of individuals in media, entertainment, and other public sectors.
Similar Image Detection
Similar Image Detection identifies if the same or visually similar images have been uploaded previously. This feature is essential for preventing duplicate content and identifying repeated uploads of harmful or inappropriate material.
How it works:
The uploaded image is analyzed and compared against a database of previously submitted images.
Image features such as patterns, colors, and textures are used to detect similarities, even if the image has been altered (e.g., resized, cropped, or slightly edited).
If a match or near-match is found, the system can flag the image for review or automatically take action based on predefined rules.
AI-Generated Image Detection
Lasso Moderation can detect whether an image has been generated using artificial intelligence. This is useful for platforms that need to monitor and filter AI-generated content, particularly in contexts where such images may mislead users or violate platform policies.
These image moderation tools can be used in combination or individually based on your platform's specific needs. Lasso Moderation offers comprehensive, customizable, and accurate image analysis to keep your content safe and compliant.
Last updated