Value:0 - 100

pixolution flow calculates a score for every document indicating the visual relevance of a document to the one referenced by rank.by and so does Solr itself for textual relevance. Both scores are combined, so that the resulting score represents visual and textual similarity.

You can control whether to focus on textually or visually similar results by setting the rank.conf.weight parameter:

  • Weight 0% : only the Solr scoring is used. No visual aspect is part of the result.
  • Weight 1-99% : textual and visual relevance are combined according to the specific weighting.
  • Weight 100% : only the visual relevance is processed. No textual aspect is part of the result (except the filter criteria of the query).

Example search for dog images looking similar to image with id 123 and a scoring focus on visual similarity: