TODO.md

# TODO
* Add description to prediction.
  * [x] Which class, is it consistent ?
  * If consistent, which properties matched ?
  * If not, which properties were an issue ?
  * For each property, sentence per target class, sentence to explain the threshold...
* Interface with Python and Ontology
  * Extract information to json file from Python
  * Read and use information in JS
* Display which target class / property were an issue with color coding or some other form (warning sign ?).
* Propose two displays, one with tabs, the other without tabs.
* Create an index page, with all the images from validation dataset, if the prediction is correct or consistent and be able to click it to go to the main explanation interface.
* Write explanation generation pattern for each property in the ontology and use it to generate explanation.