Updating the changelog for Bob's release

Please update log changes for the packages that you have committed to. The procedure is as following:

1. For a given package, for each tag (version), summarize the commits (lines start 
with '-' character) into several descriptive lines. Make sure that the new summary 
lines start with `*` character (so, they become tag descriptions).

2. The last tag name is called `patch` by default. This indicates that a patch 
version of this package will be automatically updated during the next tag/release. 
You can change `patch` to `minor` or `major`, and the package will be then 
tagged/released with either minor or major version bump. 

3. Once all commits were changed to corresponding tag descriptions (no more lines 
start with `-` characters), this package is ready for release and you can continue 
to another package in the changelog.

release/changelog_since_last_release.txt

@@ -4,12 +4,13 @@
   * v2.1.4 (Sep 22, 2017 11:34)
      * port to new CI/CD
   * v2.1.5 (Sep 22, 2017 13:58)
-     * no signficant changes
+     * no significant changes
   * v2.1.6 (Feb 27, 2018 16:29)
      * Enables built-in conda packaging
      * Provides user/development guides
      * Provides an alternative for gdb (based on lldb) for MacOS
      * Improve handling of environments for Python 3.6 (issue #24)
+  * none
 * bob.extension
   * major
      * Big changes to the documentation of this package: divided guide in
@@ -48,32 +49,12 @@
      * trust numpy in finding blas/lapack. Fixes bob.math#14
 * bob.measure
   * major
-     - Fixed curve to show values between [0,1]
-     - Fixed issue with FAR and FRR values being exactly 1 or 0, respectively
-     - Update apply_threshold.py
-     - Provide a test case for negative thresholds
-     - remove extra import
-     - Added a warning before returning NaN
-     - Changed test case for different corner cases
-     - Add some obvious test cases
-     - Correct the threshold calculation implementation in far_threshold and…
-     - rename check_index to index
-     - Add more tests
-     - implement the tests correctly
-     - Never return NaNs for {far,frr}_threshold functions anymore
-     - Change the version too
-     - Compute roc using roc_for_far internally
-     - Iterate only once in the sorted negatives to keep the implementation in O(N)
-     - Use epsilon (2^-52) instead of 1e-12
-     - use nexttoward instead of epsilon since epsilon is too small sometimes
-     - use C nexttoward
-     * Migrate to conda based CI
-     - use std::nextafter instead of nexttoward
-     - Add an option to capture the output in a log file
-     - Add the expected output file for tests
-     - remove left-over docs
-     - recompute far values in roc_for_far
-     - Fix the test values
+     * Fixed issue with FAR and FRR values being exactly 1 or 0
+     * Corrected the threshold calculation in far_threshold() and frr_threshold(), which now guarantees to produce at most of given FAR or FRR values.
+     * Never return NaNs for far_threshold() and frr_threshold() functions
+     * Optimized implementation of ROC for a given FAR
+     * Added an option to capture the output in a log file
+     * Updated docs and tests accordingly
 * bob.io.image
   * minor
      * Fixed issue with PGM image load when such images have extra bytes, see bob.io.image!36
@@ -158,7 +139,8 @@
   * minor
      * Migrate to conda based CI
 * bob.ip.optflow.liu
-  * patch
+  * minor
+     * Migrate to conda based CI
 * bob.ip.flandmark
   * minor
      * Migrate to conda based CI
@@ -168,22 +150,13 @@
      * Appended the idiap gpu short queue in our queue list
      * Improved python 3 compatibility (Merge gridtk!7)
      * Created the `jgen` tool (https://www.idiap.ch/software/bob/docs/bob/gridtk/master/generate.html) -- Merge gridtk!8
-     - Add sgpu to the list of the queues
-     - Submit extactly the same job several times
-     - Removed overwriting job ID when stopping jobs
-  * patch
-     - Add an option to make the aggregate output unique
-     - keep the whitespace
-     - Add tests for jgen --unique-aggregate
-     - make sure log_dir is directory is a directory
-     - improve the error message
-     * Migrate to conda based CI
-     - SGE scripts cannot be tested on the CI
-     - Fix sort errors that happen in Python 3
-     - Add a test for when groupby is used in templates
-     - Make the tests less relaxed
+     * Removed overwriting job ID when stopping jobs
+  * minor
+     * `jgen` tool: added an option `--unique-aggregate` to make the aggregate output unique 
+     * Migrate to conda based CI
+     * Improve tests and docs
 * bob.ip.qualitymeasure
-  * patch
+  * minor
      * Migrated to conda based CI
      * Updates in the README
      * Removed bootstrap-buildout functionality
@@ -322,122 +295,51 @@
   * v2.4.1 (Sep 22, 2017 14:32)
      * Packaging changes - no changes to software.
   * major
-     - Fix packaging warnings
-     - Migrate to conda based CI
-     - Overall database refactoring; Improved docs and tests
-     - Add information on the annotations
-     - Fix function comment
-     - Fix typo in guide
+     * Migrate to conda based CI
+     * Overall database refactoring; Improved docs and tests
+     * Add information on the annotations
 * bob.db.verafinger
   * v1.0.0 (Oct 11, 2017 14:24)
      * Initial release of new VERA fingervein database interface. Deprecates https://gitlab.idiap.ch/bob/bob.db.vera.
-     - Upload all metadata to our server
-     - Fix pkg_resources import
-     - Delete metadata (ship with db files)
-     - Update badges
-     - Reset initial version
-     - Reset authorship of package [ci skip]
   * major
-     - Update ignore file
-     - Fix packaging warnings
-     - Migrate to conda based CI
-     - Improved documentation; Simplified protocol files
-     - Add information on the annotations
-     - Fix function comment
-     - Fix typo in guide
-     - Unify biometric/presentation-attack interfaces
-     - Implement cropped biometric recognition protocols
-     - Updated documentation to include more image samples and information on biorec cropped protocols
+     * Migrate to conda based CI
+     * Add information on the annotations
+     * Unify biometric/presentation-attack interfaces
+     * Implement cropped biometric recognition protocols
+     * Improved documentation and simplified protocol files
 * bob.db.fv3d
   * v1.0.0 (Oct 11, 2017 15:42)
-     - Update badges
-     - Remove extra import
-     - Distribute metadata from webserver
-  * major
-     - Fix packaging warnings
-     - Use model name for scoring
-     - Fixes to model.id -> model.name as model identifier
-     - Migrate to conda based CI
+     * Distribute metadata from webserver
+  * minor
+     * Fixes to model.id -> model.name as model identifier
+     * Migrate to conda based CI
 * bob.db.hkpu
   * major
-     - Initial commit
-     - Incorporate model-id on File class
-     - Remove unique finger name attribute as that does not match model-id
-     - Fix tests
-     - Remove tests for finger-name (removed)
-     - Add explanation for returning just one single plane from original image
-     - Fix comment
-     - Set private URLs for docs [ci skip]
-     - Fix docs
-     - Add functions to detect and load RoI annotations
-     - Change protocol A to have multiple models instead of multiple samples per model
-     - Update filelist
-     - Update HKPU filelist with new annotations
-     - Add test for annotations
-     - Fix function comment
-     - Add information on the annotations
-     - Fix typo in guide
-     - Port to new CI (closes bob.db.hkpu#1)
-     - Fix build deps
+     * Initial release of Hong Kong Polytechnic University Finger Image Database
+     * Added docs and tests
+     * Update HKPU filelist with new annotations
+     * Migrate to conda based CI (closes bob.db.hkpu#1)
 * bob.db.thufvdt
   * major
-     - Initial commit
-     - Add protocol files
-     - Set protocol p3 to match paper descriptions
-     - List protocol files
-     - Fix sql3 file creation
-     - Finalize tests
-     - Import correct module on docs
-     - Fix short descr
-     - Centralize links
-     - Add functions to detect and load RoI annotations
-     - Change db name to thufvdt (simpler)
-     - Updated ignore file
-     - Update metadata contents
-     - Add symlinks to metadata
-     - Add annotation check tests
-     - Add annotation section
-     - Fix typo in func doc
-     - Port to new CI (closes bob.db.thufvdt#1)
-     - Fix build deps
-     - Fix build deps
-     - Pin compatible pillow versions
-     - Fix use of pin_compatible
+     * Initial release of Tsinghua University Finger Vein and Finger Dorsal Texture Database
+     * Added docs and tests
+     * Added protocols so they match paper descriptions
+     * Added functions to detect and load RoI annotations
+     * Migrate to conda based CI (closes bob.db.thufvdt#1)
 * bob.db.mmcbnu6k
   * major
-     - Initial commit
-     - Add protocols; Fix creation
-     - Fix tests
-     - Implement roi-based protocol; improve testing
-     - Rectify score count analysis
-     - Avoid creation of unused Finger entries on SQL database
-     - Fix comment
-     - Use model name for scoring
-     - Fixes to model.id -> model.name as model identifier
-     - Return any finger name (all are the same) given the first found model
-     - Fix finger name look-up
-     - Fix function comment
-     - Add information on the annotations
-     - Update file listing to include annotations
-     - Add annotation tests
-     - Add warning about annotations
-     - Include annotations with metadata
-     - Test annotations
-     - Port to new CI (closes bob.db.mmcbnu6k#1)
+     * Initial release of Chonbuk Nation University Fingervein Database
+     * Add protocols and annotations
+     * Fixes to model.id -> model.name as model identifier
+     * Added docs and tests
+     * Migrate to conda based CI (closes bob.db.mmcbnu6k#1)
 * bob.db.hmtvein
   * major
-     - Initial commit
-     - Correct database creation
-     - Fix test
-     - Fix references in documentation
-     - Use model name for scoring
-     - Fixes to model.id -> model.name as model identifier
-     - Return any finger name (all are the same) given the first found model
-     - Fix finger name look-up
-     - Add annotations
-     - Add information on the annotations
-     - Fix function comment
-     - Port to new CI (closes bob.db.hmtvein#1)
+     * Initial release of Homologous Multimodal Traits (Fingervein) Database
+     * Fixes to model.id -> model.name as model identifier
+     * Return any finger name (all are the same) given the first found model
+     * Added annotations, documentation, and tests
+     * Migrate to conda based CI (closes bob.db.hmtvein#1)
 * bob.db.voicepa
   * v1.0.2 (Sep 22, 2017 14:34)
   * minor
@@ -462,42 +364,15 @@
   * v3.3.0 (Sep 22, 2017 13:41)
      * bob.bio.base!20 Docs url fix
      * bob.bio.base!19 Python36 & docs fixes
-     * bob.bio.base!18 Removed`Algorithm.read_probe`
-  * patch
-     - Add as_array to frame containers
-     - Fixed tuple indexing bug in youtube db load function
-     - Added unit test for the load method of youtube database
-     - Updated manifest to use jpg files for tests
-     - Fixed the unit test connecting it with low level db interface]
-     - Fixed the unit test of youtube load method
-     - Make the original_directory and annotation_directory a property
-     - move the _db initialization before calling super
-     - pass kwargs in the base file class
-     - Migrate to conda based CI
-     - [conda] Added test requirements
-     - Fixed the index.rst to match joint docs requirements
-     - Added quality_assessment_function to preprocessor Wrapper
-     - Fixed the typo
-     - Addressing missing frame annotations
-     - Fixed dict.has_key issue for Python 3
-     - Reverted quality assessment function
-     - Changed missing annotations conditional check + unit tests
-     - Moved condition to one line, removed empty dictionary check
-     - Fixes bob.bio.base#11
-     - Returning none insted of empty FrCont in preproc Wrapper
-     - Make missing annotations None instead of empty dict
-     - Add a failsafevideo wrapper
-     - use ordered dict
-     - Add a video annotator
-     - Finalize video annotators and add docs and tests
-     - Fix import order
-     - use six to check for string types
-     - Account for empty frame containers too
-     - Account for failed extractor
-     - remove unused import thanks to my linter and the way I imported it
-     - mention min_face_size_validator for a possible input
-     - Validators should handle empty annotations too
-     - Fix sphinx warning
+     * bob.bio.base!18 Removed `Algorithm.read_probe`
+  * major
+     * Add as_array to frame containers
+     * Make the original_directory and annotation_directory a property
+     * Fixed tuple indexing bug in youtube db load function and added tests
+     * Updated docs and tests
+     * Fixed bob.bio.base#11
+     * Added video annotator, updated documentation accordingly
+     * Migrate to conda based CI
 * bob.bio.gmm
   * v3.1.0 (Sep 22, 2017 14:05)
      * bob.bio.gmm!10 Removed Algorithm.read_probe method, since this is already solved via bob.bio.base
@@ -509,69 +384,18 @@
      * Propagated --allow-missing-files to the UBM training
      * Fixed issue bob.bio.gmm#22
 * bob.bio.face
-  * patch
-     - First commit
-     - Added ignore
-     - Added ignore
-     - IJBA
-     - Updated experiments
-     - Updated environment
-     - [ci] Added the project in the ci
-     - Updated some baselines
-     - Updated some tables
-     - UPdate results
-     - Updated baselines
-     - Updated some baselines
-     - Updated baselines
-     - Updated baselines
-     - Reformulated the package
-     - Reformulated the package
-     - Updated baseline script
-     - Updated baselines
-     - Updated baselines
-     - Updated baselines
-     - Updated some paths
-     - Tensorflow hack
-     - Fixed issues with baseline script
-     - Added IJBB
-     - Cleaning up
-     - fixed IJBB
-     - Cleaning up the module
-     - [sphinx] Updated docs
-     - Merging
-     - [sphinx] Updates in the documentation
-     - Appended more baselines
-     - Appended more baselines
-     - New baselines
-     - Update MSCELEB baseline
-     - Updated baselines
-     - Updated baselines
-     - Updated CASIA
-     - Updated CASIA
-     - Updated baselines
-     - Updated MSCeleb
-     - Updated MSCeleb
-     - Updated MSCeleb
-     - Updated MSCeleb
-     - Added inceptionv1
-     - Added inceptionv1
-     - Updated baselines
-     - Inception v1
-     - Inception v1
-     - Inception v1
-     - Inception v1
-     - Inception v1
-     - Updated baselines
-     - Updated baselines
-     - updated batchnorm
-     - Added new baseline
-     - BAselines
-     - New center loss
-     - Update
-     - Update
-     - New baseline mechanism
-     - [ci] Removed mac builds
-     - [ci] Update conda
+  * v3.3.0 (Sep 22, 2017 13:17)
+     * bob.bio.face!78 Removed Algorithm.read_probe method, since this is already solved via bob.bio.base
+     * bob.bio.face!33 improve the score files by adding attack type information
+     * [ci] Droped python 3.4 support and added python 3.6
+     * bob.bio.face!30 Updated baseline script to search for executables
+  * major
+     * Implemented IJBA customized plot
+     * Make the original_directory and annotation_directory a property
+     * Created IJBB high level interface
+     * Added script to display face annotations provided by a database. 
+     * Migrate to conda based CI
+     * improve docs and tests
 * bob.bio.spear
   * v3.1.1 (Sep 22, 2017 14:48)
      * bob.bio.spear!36 Docs url fixes
@@ -622,7 +446,7 @@
      * Remove Oleg's implementation of rotation-invariant matching
   * v1.0.1 (Oct 20, 2017 16:04)
      * Update `MANIFEST.in` to include necessary package files
-  * patch
+  * minor
      * Ported to new CI
      * Updated utfvp interface
      * Fixed the index.rst to match joint docs requirements
@@ -671,38 +495,13 @@
      * Refactoring: removed unused VideoDataLoader
      * Refactoring: created FaceCropAlign, replaces ImageFaceCrop + VideFaceCrop using Wrapper
 * bob.db.oulunpu
-  * patch
-     - Initial commit
-     - Implement frames
-     - add ci
-     - update license
-     - Improve readme
-     - Add a config file
-     - Correct frame size in docs
-     - include list files in the package as well
-     - Make sure all_files is working
-     - models depends on protocol
-     - Add frame_size methods
-     - Adopt to the new database
-     - Use bob.pad instead of bob.dap
-     - fixes
-     - fixes
-     - Fix entry point
-     - Use bob.extension.rc
-     - Add bob.bio.config entry point
-     - Fix the config file
-     - Add conda packaging
-     - Updates CI file
-     - Add annotation support and reorganize frame reading
-     - fix sphinx warnings
+  * major
+     * Initial release of OULU-NPU Database
+     * Added protocols, documentation, and tests
+     * Add annotation support and reorganize frame reading
+     * Migrated to conda based CI
 * bob.db.uvad
-  * patch
-     - Initial commit
-     - Fix entry point
-     - Use bob.extension.rc
-     - Adopt the tests as well
-     - improve tests
-     - Fix cropping
-     - Add bob.bio.config entry point
-     - Add conda packaging
-     - Updates CI file
+  * major
+     * Initial release of Unicamp Video-Attack Database
+     * Added protocols, documentation, and tests
+     * Migrated to conda based CI