What is ImageSnippets?
ImageSnippets is a Linked Data Annotation System.
Anyone who archives, curates or shares images on the web can use ImageSnippets to add 5-star linked open data descriptions to your images.
You can also search, manage, share and publish images with the same data.
Use ImageSnippets on your own, or allow us to help you annotate your images with industry standard rich linked open data (LOD), embedded metadata and alt-text descriptions for accessiblity.
The system also preserves schema.org, JSON-LD and RDFa with your images.
 |
 |
 |
 |
Optimizing Image Description
Captioning, Classifying, Labeling, Knowledge Engineering
Our image annotation and publishing tools and services are human and data-centric. They are augmented by AI techniques to speed up the annotation process while allowing the user to have control over context and precision. Linked data connects to precise entities from Wikidata, DBpedia and other. Create your own custom vocabularies where necessary.
Please email us to discuss your image description/metadata needs: info@metadata.rocks
Graphs of many flavors: Personal Knowledge, Scene Graphs, Multi-Modal Graphs, Domain Specific Graphs
The flexibility of our triple-tagging techniques allows a variety of graphing tasks to be accomplished easily. Graphs made from image descriptions become multi-modal representations of domain knowledge interwoven with the images.
Ontology Engineering
Images provide excellent starting points for identifying important concepts and grounding those concepts. Visually identifying ground truth is especially helpful in collaborative knowledge engineering environments.Accessibility
Accessibility ---- We provide solutions for annotating images with WCAG compliant alt-text. Our system ensures that descriptive data travels with images using industry-standard IPTC metadata. All of our systems are managed by, with and for humans who are at the center of our computing solutions.
Transforming Classifiers into Knowledge Graphs
Machine Learning often classifies images with lists of superfluous and unrelated classifiers - often with similar confidence on conflicting concepts. Classifiers transformed into much more precise, structured graph data through named entity recognition techniques combined with inference allows the graph data to then become useful for a variety of tasks: identifying disagreement in ground truth concepts, active learning for machines, search, publishing and troubleshooting knowledge graph modeling.
AI Blindspot Alignment
Humans and machines BOTH make mistakes, but, they can also complement each other. Machines can spot things that humans miss and humans understand things that machines haven't learned yet. To fix these issues, our system allows ML detectors to be plugged into real-world environments so blindspots can be identified, graphed and resolved, while also allowing humans to correct their own blindspots using object detection and machine interpretation.
Our Research
Our research comes from over a decade of using the ImageSnippets system to construct graph data about domain-specific image collections.
Learn more about our team here: The Metadata Authoring Systems Team