With Visual UGC's Advanced Visual Recognition, you will have access to a number of custom models that are trained to identify specific subcategories of objects.
These models are more accurate at detecting the content they are trained on than a general model. For best results, you should apply models to content you believe will have a high probability of containing relevant objects.
To choose which models to apply, navigate to Settings > Visual Recognition > Advanced Visual Recognition Models and select the checkboxes to choose which models you would like Visual Recognition to pass your content through for analysis.
What does each Model do?
The ‘General’ model recognizes over 11,000 different concepts including objects, themes, moods, and more. When using the General model you can choose to exclude human characteristics concepts such as demographic, ethnicity, gender, etc. if that doesn't comply with your organization's policies.
The ‘NSFW’ (Not Safe For Work) model determines the likelihood that an image contains nudity. It is a great model to apply if you have a high volume Term such as a public hashtag.
The ‘Color’ model returns density values for dominant colors present in images. The color model is great for filtering content that fits in a particular color scheme.
The ‘Travel’ model is an industry specific model that recognizes features of residential, hotel, and travel-related properties.
The ‘Apparel’ model recognizes more than 100 fashion-related concepts including clothing and accessories.
Want to uncover the celebrities in your content? The ‘Celebrity’ model identifies the face(s) of over 10,000 recognized celebrities.
The ‘Food’ model recognizes more than 1,000 individual food items as well as identifying individual ingredients.
The ‘Logo’ model is trained to identify the logos of over 500 recognized brand names.
Want to add your logo to the 500? Custom models can be trained for your brand - submit a request to ask us how.
This model is similar to NSFW but with a broader scope. It identifies concepts such as gore, drugs, explicit nudity, or suggestive nudity.
As you would expect, the ‘Wedding’ model recognizes concepts related to weddings - over 400 of them in fact.
The 'demographics' model predicts the age, gender, and cultural appearance of detected faces.
Texture & Patterns
The 'Texture & Patterns' model identifies textures and patterns within an image.
Social Popular Shots
The 'Social Popular Shots' classifies popular Social Post types (ie. Selfies, Landscape)
The 'Room Model' identifies interior rooms within houses and buildings.
For any further questions or queries, please submit a request and our support team will get back to you.