Tags give the ability to mark specific points in history as being important
- Significant changes:
- Added new plotting and evaluation scripts. Refer to docs.
- Detailed Changes:
- !52 Change assert to assert_click_runner_result
- !50 Add new classification algorithms: As mentioned here !33 here is the new branch
- !51 Fix broken commands cref bob.extension!86
- !48 Various fix: Requires bob.measure!82 Similar to bob.measure!82 for PAD
- !45 Allow PAD filelist database to be used in vulnerability experiments
- !47 Remove bob vuln metrics and evaluate commands: since they were not well defined and we do not know what should be in there. rename --hlines-at to --fnmr in bob vuln roc,det commands small nit-pick fixes overall
- !43 Finalization of plots: Fixes #22. Requires merge bob.measure!65. Requires bob.measure!67
- !46 Add a command for multi protocol (N-fold cross validation) analysis: Depends on bob.measure!79
- !44 Remove the grid_search.py entrypoint: Fixes #23
- !41 Set of click commands for pad: Provide commands: bob pad metrics bob pad epc bob pad epsc bob pad gen
- !42 Set of click commands for pad: Provide commands: bob pad metrics bob pad epc bob pad epsc bob pad gen
- added MLP and LDA classifiers for PAD
- Migrate to conda based CI
- Added --allow-missing-files option
- Improved documentation
- Moved from bob.pad.face SVM (bob implementation), one class GMM (scikit-learn) and logistic regression (scikit-learn) algorithms
- Cleaned up and improved code. Added tests.
- Replace the old PAD ISO metrics class with a script
- Added clients option to FileListDB driver
- Avoids feature blow-up when std==0.
- Added predictions algorithm and padfile to label.
- Added support for annotations files
- Added support for configuration files (describe experiment in config file and run it with
- Added support for filelist-based databases