Classification models
Machine learning models that analyze the content of an image and assign it to predefined categories or labels
Classification models in DAM
Machine learning models, particularly classification models, play a crucial role in digital asset management. Classification models are algorithms that learn patterns and relationships in data to assign predefined categories or labels to new, unseen data. They can be used to categorize and organize assets, such as images, videos, or documents, based on their content or characteristics.
Classification models are typically trained on labeled datasets, where each data instance is associated with a known category or label. During the training process, the model learns to recognize patterns and features in the input data that differentiate one category from another. Once trained, the model can classify new, unseen data based on the learned patterns.
Auto-tagging
Automatic assignment of relevant tags and keywords to the input image
Overview
Harnessing the power of visual assets is crucial for businesses across various industries. However, as digital libraries grow exponentially, manually assigning tags and keywords to images becomes a daunting and time-consuming task. That's where automatic tagging can help.
The Image tagging ML model is a powerful machine learning algorithm integrated into our service. It automatically analyzes the pixel content of images, extracts their features, and assigns relevant tags or keywords to facilitate improved organization and utilization of visual assets. With this model, businesses can unlock a new level of efficiency and productivity.
The process of AI tagging involves object detection techniques that detect objects, scenes, and concepts within the images. Based on this analysis, appropriate tags are assigned to each image, enabling proper asset management and content discovery.
Use cases
The AI tagging model can provide support for a wide range of typical use cases, including:
Efficient organization - Automatically assigned tags enable efficient organization of visual assets by categorizing them based on specific objects, scenes or concepts. Users can easily navigate through their image library and quickly locate the desired assets.
Accelerated search - Users can retrieve images based on generated tags or keywords. This accelerates the search process, making it effortless to find the required images in large collections.
Content Recommendation - Visually similar assets can be identified based on shared tags. Users can discover related content and leverage the full potential of their visual assets.
Visual content analysis - The model can aid in understanding visual trends, analyzing customer preferences and performing visual data analytics.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
API response
API response
Brand detect
An ML model that detects logos of popular brands, visible in the input image
Overview
In today's digital landscape, brands are a vital aspect of any business. A brand logo serves as a symbol of identity, recognition, and trust. Managing and protecting brand assets is essential for maintaining brand consistency, ensuring compliance, and maximizing brand visibility. Our model is designed to streamline this process by automating the detection and categorization of brand logos within your digital assets, facilitating effective brand asset management.
The Brand recognizer is powered by state-of-the-art machine learning algorithms, and trained on vast datasets containing diverse brands from a wide selection of industries. This enables the model to accurately recognize and detect brand logos with varying sizes, orientations, and backgrounds within the provided images.
The model supports a vast number of popular brands and enables efficient brand analytics and monitoring, providing valuable insights and efficient brand management within our service.
Use cases
The following are some use cases, suitable for our Brand recognition model:
Brand monitoring and compliance - By identifying brand logos in user-generated content, the model assists in brand monitoring, ensuring compliance with brand guidelines and protecting brand integrity.
Brand analytics - The model provides insights into brand visibility and presence by analyzing the frequency and distribution of detected brand logos across digital assets. This information can be used for brand performance evaluation and marketing strategies.
Competition Analysis - By recognizing competitor brand logos within the digital assets, the model aids in conducting competitive analysis, identifying market trends, and evaluating brand positioning.
Image Retrieval - The Brand Recognizer enables efficient searching and retrieval of digital assets by logo-based queries. Users can locate specific assets based on recognized brands.
Digital rights management - Copyright and licensing agreements can be enforced with automatic logo detection and tracking. This ensures compliance with intellectual property regulations.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
Output preview
Output preview
API response
API response
Dominant color extraction
An ML model that extracts the most prominent colors in an image
Overview
The model analyzes the colors present in the input image and determines if they fall into several predefined ranges. It then compares the number of pixels in those ranges and outputs color names that were detected as visually dominant.
The pixels of the image are examined to identify the most prominent colors present. Multiple factors such as the distribution and frequency of different color values are taken into account to determine the dominant colors.
To find the color of a pixel, all values in the image are converted to an alternative color representation that is designed to be more closely aligned with the way human vision perceives colors. This enures high-quality results.
Use cases
Some typical use cases for this model include:
Visual search and filtering - Users can search for images based on specific dominant color criteria. For example, they can search for images dominated by a particular color to find assets that align with a specific theme or aesthetic.
Color-based categorization - The feature facilitates the automatic categorization of images based on their dominant colors. This allows for easier location and grouping of images with similar color schemes.
Branding Enhancement - Teams can utilize the dominant color information to ensure consistent branding and visual coherence. They can search for images that match their brand identity and integrate them into their campaigns.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
Faces
Face analysis
A set of ML models that accurately detect human faces and predict crucial characteristics like facial landmarks, expression, ethnicity, age, and gender
Face analysis
A set of ML models that accurately detect human faces and predict crucial characteristics like facial landmarks, expression, ethnicity, age, and gender
Overview
The Face analyzer consists of several cutting-edge ML models. Its primary function is to detect visible human faces in images and predict some facial characteristics that are deemed important.
Leveraging state-of-the-art deep learning algorithms and neural networks, the Analyzer accurately identifies and analyzes faces, extracting the following information for each face:
position in the image (bounding box);
facial landmarks (coordinates of points that map to specific facial structures on the face);
expression classification (happy, angry, sad, etc.);
ethnicity classification;
age estimation;
gender classification.
Use cases
Use cases for automatic facial analysis include:
Image tagging and organization - Automatic facial analysis enables users to easily categorize and index images based on expressions, ethnicities, age groups and genders. This streamlines content management, making it easier to locate specific images for various purposes.
Inclusive representation - The Face Analyzer can facilitate inclusive representation in media content. By analyzing facial ethnicities and genders, content creators can ensure diversity and cultural representation in their visual assets, thus promoting inclusivity.
Search Optimization - Automatic tagging and categorization of images based on facial characteristics allow users to find specific faces, expressions, or ethnicities with ease.
API endpoints
An up-to-date reference with all API endpoints is available here:
Scaleflex API for Digital Asset Management (DAM), Visual AI and Media OptimizationScaleflex API
Example API responses
Output preview
Output preview
API response
API response
Face clustering
A system that detects human faces, extracts feature vectors and clusters similar faces to efficiently group images based on the individuals present in them
Overview
The Face Clustering ML model is an advanced component offered as a part of our service. With its powerful facial detection and feature extraction capabilities, this model fundamentally changes the way images are organized and grouped based on the individuals present in them.
The model accurately detects visible human faces within images, regardless of their positions or orientations. This crucial first step ensures that all relevant faces are identified for subsequent clustering.
For each detected face, the model extracts a high-dimensional feature vector, representing the unique facial characteristics and attributes of that individual. These feature vectors capture critical facial traits while discarding irrelevant information, making them ideal for face-based comparisons.
Leveraging advanced machine learning algorithms, the model clusters all faces (and the respective images they are part of) based on the similarity of their feature vectors. Faces with similar features are grouped, enabling the automatic identification of distinct individuals in image collections.
The face clustering model can categorize faces without the need for explicit training data on individual identities. This flexibility allows the model to adapt to various datasets and expand its clustering capabilities.
Use cases
Automatic face clustering can have a significant impact on the following use cases:
Event photography management - Event photographs often contain multiple different individuals. The model simplifies the organization of these images by automatically grouping them based on the distinct people present. This streamlines the curation and delivery of photo collections.
Content organization - Automatic face clustering can make it easier for users to navigate and discover images related to specific individuals.
Photo collections - In personal photo libraries, the face clustering model assists in creating dedicated albums or collections for family members and friends, thus saving users valuable time by automatically organizing photos by individual identities.
API endpoints
An up-to-date reference with all API endpoints is available here:
Examples
Input images
Detected faces
Created clusters
Image quality
A ML model that evaluates the technical quality of images by analyzing their visual attributes
Overview
The quality of digital content is a critical factor in delivering satisfactory user experience and achieving branding goals. Our Image quality assessment (IQA) model is a machine learning solution designed to evaluate the visual quality of images.
When an image is fed into the system, the model processes it to assess various quality attributes and generates a quality score that reflects a weighted combination of those attributes, providing a precise assessment of the image's technical quality.
The attributes include multiple factors such as sharpness, blurriness, noise, dynamic range, contrast, distortion, etc. By providing a thorough image quality assessment, the model greatly enhances the management of digital assets and offers indispensable aid in content selection and optimization.
Typical use cases
Image quality assessment finds application in various domains and industries where image quality is a top priority:
Stock photography - Stock image providers can use the model to rate and categorize the images in their libraries, thus helping customers find ones with superior quality for their projects.
E-commerce - E-commerce platforms can ensure product images meet quality standards, improving the visual appeal of product listings and driving sales.
Publishing - Media outlets and publishing companies can automate the assessment of images for articles so their publications contain high-quality assets.
Advertising - Advertising agencies can use more easily evaluate the technical quality of campaign visuals, helping maintain brand consistency and professionalism.
Digital marketing - Marketers can optimize their social media and online content for greater impact and engagement.
API endpoints
Information about the specific API endpoints is available in an always up-to-date documentation, that can be accessed via the following link:
There, you can find detailed information about the API endpoints, together with all required request parameters, so you know how to interact with them.
Example API responses
OCR
A machine-learning algorithm that utilizes optical character recognition techniques to accurately identify and extract text from images or scanned documents
Overview
The OCR ML model is an integral component of our service, providing robust optical character recognition capabilities with support for multiple languages.
The model is designed to convert text present in images or scanned documents into editable and searchable data. It leverages the power of machine learning techniques to accurately recognize characters and words, enabling efficient text extraction and analysis.
It examines the input, tries to detect any text fragments present in the image and recognizes the characters in those fragments according to the specified language. Detected fragments with a good enough confidence level are returned as text strings.
Use cases
The OCR service can be useful for multiple use cases, including:
Text extraction and indexing - The model extracts text from images or scanned documents, enabling efficient indexing and searching of digital assets. Users can find images or documents based on specific keywords or phrases mentioned within the text content.
Document digitization - Important information can be preserved by converting physical documents into digital formats.
Language translation - Combined with language translation capabilities, OCR can facilitate multilingual asset management, enabling users to search and translate text in various languages.
Improved accessibility - Converting text within images or scanned documents into machine-readable format enhances accessibility for visually impaired individuals, enabling screen readers or assistive technologies to interpret the content.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
Number Plate recognition
An ML model that detects and recognizes vehicle registration plate numbers
Overview
The Number plate recognizer is a machine-learning model that detects and recognizes vehicle registration plate numbers in images.
It is developed and trained to detect vehicle license plates, recognize the characters on the plate, and return them as text strings. The model takes in an input image containing a vehicle and outputs the alphanumeric number present on the vehicle's registration plate in text form.
The process of license plate recognition consists of the following steps:
License plate detection;
Pre-processing the resulting image from Step 1 (warping, deskewing) to prepare it for optical character recognition (OCR);
Passing the image from Step 2 through an OCR engine and receiving the recognized characters.
Use cases
Some typical use cases for this model include:
Efficient vehicle cataloging - Vehicles can automatically be cataloged in a database. When new vehicle images are uploaded, the model can extract the registration plate numbers and associate them with the corresponding vehicle records, allowing for quick and accurate identification and retrieval of vehicle assets.
Smart searching - Users can leverage plate number recognition to perform targeted searches and easily locate specific vehicles, thus saving time and effort.
Automated metadata - Vehicle assets can be tagged with relevant metadata based on their registration plate numbers. This metadata may include vehicle make, model, year, and other information associated with the recognized plate.
Regulatory compliance - The model can aid in ensuring compliance with legal requirements related to vehicle registration and documentation.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
Product type
An ML model classifying an image as pure product image or application image, showing the product in use
Overview
The Product type classifier is a machine learning model, specifically designed to classify images into two categories:
Pure product images (packshot) typically show the product by itself in a clear and uncluttered way. The goal in this case is to focus on and highlight the product features and design.
Product application Images that show the product in use or context, often in a real-world setting. These images are designed to illustrate how the product can be operated, its potential benefits, and how it fits into a larger system or environment.
The model is particularly useful in the context of online stores and businesses dealing with large collections of product images. By leveraging its power, businesses can streamline their digital asset management workflows, organize image libraries, and deliver personalized user experiences.
Typical use cases
Some typical use cases include:
Image library organization - Online stores often have vast libraries of product images that need to be effectively organized and categorized. The classifier can automatically sort images based on their content, facilitating easy retrieval and management.
Search enhancement - The model allows for more accurate search and filtering capabilities. Users can easily find images that showcase the product itself or images that demonstrate its usage in real-world contexts.
Content Curation - The classifier aids in curating image galleries and content showcases by providing an automated classification of images. It aids in showcasing product images in relevant contexts, ensuring a coherent and focused presentation.
Product Recommendations - By understanding whether an image focuses on the product or its application, businesses can enhance their recommendation algorithms. The classifier can be used to tailor recommendations based on the user's preference for pure product images or application images.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
Application image
API response
Property classification
Precise classification of property-related images across various industries.
Overview
Efficient categorization of digital content is essential for organizing and retrieving assets effectively. The Property classification model is an advanced machine learning (ML) solution designed to identify and classify the type of real estate property depicted within an image.
The model analyzes the input image and determines if the displayed realty falls into one of the following property classes:
Apartment
Bathroom
Bedroom
Dining
Garage
House
Industrial
Kitchen
Living
Office
The Property classification feature is based on state-of-the-art computer vision techniques and deep learning algorithms that interpret the image data and identify patterns that are indicative of the property type depicted. This allows the model to accurately predict the predefined category that the image is part of, enabling precise classification. By doing this, this model significantly enhances the management of digital assets, streamlining content organization and retrieval for various industries.
Typical use cases
Property classification finds utility in a variety of use cases, all aimed at improving digital asset management and content organization:
Real estate listings - Real estate agencies can use the model to automatically classify and organize property images for listing websites, making it easier for potential buyers or renters to find properties of interest.
Property management - Property management companies can organize images of rental properties, simplifying property management and tenant communication.
Interior design - Interior design platforms can categorize images of furniture and home decoration products, thus enhancing the online shopping experience for customers.
Hospitality industry - The model can help hotels and hospitality businesses sort images of their rooms, facilities and dining areas for marketing and reservation platforms.
E-commerce product classification - E-commerce platforms can use the model to automatically categorize products based on the type of setting they are intended for (e.g., kitchen appliances, bedroom furniture).
API endpoints
Information about the specific API endpoints is available in an always up-to-date documentation, that can be accessed via the following link:
There, you can find detailed information about the API endpoints, together with all required request parameters, so you know how to interact with them.
Example API responses
Scene Classification
An ML model that accurately classifies images, offering appropriate scene categories and attributes
Overview
The Scene classifier is a cutting-edge artificial intelligence system integrated into our service. Its primary function is to analyze and classify images based on their environmental context and distinguish between indoor and outdoor scenes. Leveraging the power of deep learning and neural networks, this model can accurately determine the setting of an image and provide insights about various scene categories, that the image has been determined to be part of.
The model offers a rich set of scene categories and attributes to help users understand the context of their images better. Some of the scene categories include landscapes, beaches, mountains, forests, interiors, exteriors, etc. It can also identify specific attributes like natural lighting, artificial lighting, crowded spaces, open spaces, urban settings, rural settings and more.
Typical Use Cases
Scene classification can be appropriate in the following use cases:
Content organization - The Scene classifier plays a vital role in automatically organizing large collections of images. By accurately determining whether an image is taken indoors or outdoors, users can easily filter and group their assets based on environmental context, streamlining content management and retrieval.
Marketing campaigns - Marketing teams can sort and analyze images that are relevant to specific campaigns. For instance, for an outdoor adventure campaign, the system can help select images of landscapes, mountains, and beaches, thus ensuring consistent and fitting visuals.
Search optimization - Scene classification adds a new layer of metadata to the images. This enhances search capabilities, enabling efficient searching for specific scenes or environments.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
API response
API response
Sport Classification
An ML model that identifies sport activities depicted in images, empowering diverse industries with precise content organization and targeted applications
Overview
The Sport classifier is a state-of-the-art artificial intelligence service. Its primary function is to analyze and classify images based on the sports activities they depict. Utilizing cutting-edge deep learning techniques and neural networks, this model can accurately identify various sports and provide valuable insights for content organization and a wide range of applications.
The model offers a diverse set of sports categories, including but not limited to football, basketball, tennis, swimming, gymnastics, baseball and many more. It ensures precise tagging and categorization of sports-related images, making it an indispensable tool for media agencies and content managers.
Typical Use Cases
Sports classification can be applied in multiple domains, including:
Sports media coverage - By quickly identifying the sport depicted in an image, journalists and editors can efficiently organize their image archives and provide real-time coverage during sporting events, enhancing storytelling and enriching their publications.
Sports analytics - The model can aid in automating the process of collecting data by categorizing images from games and training sessions.
Sports product marketing - Sports brands and retailers can utilize the model to optimize their marketing efforts. Sorting images of athletes engaging in specific sports can aid in the creation of targeted campaigns.
Search enhancement - Precise sport categorization can optimize the process of searching and filtering assets, enabling users to find relevant images for their projects with ease.
API endpoints
An up-to-date reference with all API endpoints is available here:
Example API responses
API response
API response