Release 1.4.3

New Features

ImageUpdateFactory support for multicore configurations

Instead of using one processor configuration for all registered cores as in the past, the ImageUpdateFactory has now support for loading different core configurations. Each core can now have a different processor configuration. This way you can configure different bandwidth limitations or fields for each core.

rank.by feature

You can do a visual similarity search with raw pixolution feature data as input. This functionality was implemented in former versions as well, but was not officially supported and documented. Please refer to the current API for details.

Optimizations

Slow search performance when lots of documents do not have a feature

During a visual scoring process a WARNING will be logged for each document that does not have a feature but is part of the scoring. If you have a large image collection and many documents do not have a feature, this may result in a perfomance loss since the warning message must be created and logged. The message will now only be created if the log level is set accordingly. If you are aware of having documents without a feature and cannot change this, you may set the logging level of the class WrappedScorer to a higher level (ERROR). Please refer to the Tips & Tricks sections of the current API.

Bug Fixes

Nullpointer Exception in distributed query when documents without feature

Due to faulty initialisation of internal params, Nullpointer exceptions could be thrown when a distributed pixolution query was triggered and one or more documents in the index did not contain a feature but were part of the search. This bug is now fixed.