| One line summary
|| Tool to extract image visualisations, map to RDF for graph analysis
| Detailed description
|| tiff2RDF runs Sanselan, FITS, ImageMagick on a set of images to extract technical metadata. The fields are mapped to RDF which relates image files to specific field values:
- get file list
- on each file, run:
- Sanselan (Java image library)
- convert outputs to RDF (values of fields become resources, making it possible to see which images share which values, in a visualisation)
The RDF can then be loaded into:
- Welkin (visualisation tool)
Loading all fields into the visualisation makes it possible to choose a specific field and render the clustering around specific values - thereby visualising the consistency across the image set. For instance, if all the image files cluster around one node (for instance a value of 8 in the field bitsPerSample) then the set is consistent. If there are outliers which cluster around another node then they are inconsistent. It is easy to see the scale of the problem - a few outliers, the whole collection split or no cluseterng at all ...
It is also easy to inspect different aspects of the image files - a click of the mouse selects a different field to render on - making it easy to switch between analysing colours to file size to whatever ...
Future work could analyse file dates or directory structure to allow image file nodes to be distinguished visually - over time, by batch, version or format.
| Solution champion
|| Peter Cliff
| Git link
- Interesting comparitive results of extraction using different tools (please document above!) - different tools extract metadata fields differently ... More detail to follow
- Visualisation approach, excellent!
- Collection owner likes ability to quickly check consistency across a large collection
- Don't need to set spec in advance and then see what conformance is - much more flexible/agile
- Works for a non specialist
| Tool (link)
|| Historic photographic collection - consistency across time or suppliers