SMQTK v0.4.0 Release Notes¶
This is a minor release that provides various minor updates and fixes as well as a few new command-line tools and a new web service application.
Among the new tools include a couple classifier validation scripts for checking the performance of a classification algorithm fundamentally as well as against a specific test set.
A few MEMEX program specific scripts have been added in a separated directory, defining an ingestion process from an ElasticSearch instance through descritpor and hash code computation.
Finally, a new web service has been added that exposes the IQR process for external tools. The existing IQR demo web application still functions as it did before, but does not yet use this service under the hood.
Updates / New Features since v0.3.0¶
- Updated supervised classifier interface to no assume presence of a “negative” class.
- Fixed libSVM implementation train method to not assume “negative” class.
compute_many_descriptors.pymain work function into a new sub-module of SMQTK in in order to allow higher level compute function to be accessible from the SMQTK module API.
- Added function for asynchronously computing LSH codes for some number of input descriptor elements.
- Update to postgresql backend to lazy-connect during batch executions, preventing a connection from being made if nothing is being added.
- Added example of setting up a NearestNeighborServiceServer with live-reload enabled and how to add/process incremental ingests.
- Revised IqrSession class for generalized use (pruned down attributes to what is needed). Fixed IqrSearchApp due to changes.
Tools / Scripts
- Added CLI script for hash code generation and output to file. This script is primarily for support of LSHNearestNeighborIndex live-reload functionality.
- Added script for asynchronously computing classifications on descriptors in an index via a list of descriptor UUIDs.
- Added script for cross validating a classifier configuration for some truthed descriptors within an index. Can generate PR and ROC curves.
- Added some MEMEX specific scripts for processing and updating data from a known Solr index source.
- Added MEMEX-specific script for fetching image data from an ElasticSearch instance and transfering it locally.
- Added script for validating a classifier implementation with a model against a labeled set of descriptors. This script can also be used to conveniently train a classifier if it is a supervised classifier type.
- Added helper wrapper for generalized asynchronous function mapping to an input stream.
- Added helper function for loop progress reporting and timing.
- Added helper function for JSON configuration loading.
- Added helper for utilities, encapsulating standard argument parser and configuration loading/generation steps.
- Renamed “merge_config” to “merge_dict” and moved it to the smqtk.utils module level.
- Added IQR mostly-RESTful service application. Comes with companion text file outlining web API.
Fixes since v0.3.0¶
- Fixed memory implementation serialization bug.
- Fixed SkLearnBallTreeHashIndex model load/save functions to not use pickle due to save size issues. Now uses
numpy.savezinstead, providing better serialization and run time.