Tags give the ability to mark specific points in history as being important
  • v2.2.1
    Release v2.2.1
    • Python 3.7 support
  • v2.2.0
    Release v2.2.0
    • !73 Fix the broken if logic in ImageQualityMeasure
    • !74 Updated BATL HLDI: added rc, skin/non-skin annotations, multi-channel laoding functionalities
    • !75 Alignment for LightCNN: Added an alignment method, suitable for LightCNN
    • !76 remove missing bootstrap-buildout.py from MANIFEST.in
    • !79 Batl hldi update: Updted batl HLDI, and skin, non-skin annotation files.
    • !81 MC preprocessor: The pre-processor for multi-channel data, allowing to detect and crop faces, and extract patches from cropped images.
    • !80 Patch extraction from images
    • !83 Preprocessing and quality check
    • !77 Added CASIA-SURF database
    • !86 Batl loo protocols: Adds Leave one out protocols to the BATL HLDI.
    • !87 Chalearn pad: 1. Fixed small mistake in CASIA-SURF High level PAD BD 2. Added some stuff to a preprocessor (sanity check and logging)
    • !88 Added the doc on MLP training using AE embeddings
    • !89 Fixed CASIA SURF
    • !90 [database] added modification to high-level interface for CASIA-SURF
    • !91 Resolve "fix CASIA-SURF tests"
    • !95 Move bob.ip.dlib and bob.ip.facelandmarks to test dependencies
    • !96 Put bob.ip packages back as run requirements
    • !97 unit tests: Reverts the unit tests
    • !98 Add casia fasd and an optical flow extractor
  • v2.1.0
    Release v2.1.0
    • !68 Improve load_utils.py: * Plus some minor fixes to the frame-diff method
    • !67 HLDI for the CelebA database and quality estimation script: This MR contains two contributions: 1. The High Level DB interface for the CelebA database. 2. The quality assessment script + config file allowing to estimate the quality of preprocessed CelebA images.
    • !66 Added an option to exclude specific types of attacks from train set in BATL DB: This allows to exclude specific PAIs from the training set of the BATL DB. PAIs currently handled: makeup.
    • !65 Updated the HLDI of BATL DB, added FunnyEyes fix, and protocol joining test and dev sets
    • !64 Add support for external annotations in replaymobile
    • !60 Change the API of yield_faces
    • !62 Add a script for analyzing database face sizes
    • added preprocessors and extractors for pulse-based PAD.
  • v2.0.1
    Release v2.0.1
    • No changes.
  • v2.0.0
    Release v2.0.0
    • Added High Level Database Interface (HLDI) for MIFS database
    • Annotations are added for MIFS database
    • Added a VideoSparseCoding preprocessor
    • Updated the documentation on anomaly detection-based PAD using Aggregated Db
    • Local face detection option was added to VideoFaceCrop class, using dlib and mtcnn
    • Unit tests for the classes used in IQM-GMM pad algorithm
    • Added landmark detection utils
    • Algorithms/classifiers are adapted to other non-face modalities
    • New protocol for Aggregated Database: grandtest-train-eval
    • New protocol for Aggregated Database: grandtest-mobio
    • Frame Selector is now acceptable in HLDIs of all databases
    • Removed bootstrap-buildout.py functionality
    • Added A base PAD file class
    • New protocol for Aggregated Database: grandtest-train-eval-<num_train_samples>
    • bob.extension.rc used in HLDIs
    • Migrated to conda-based CI
    • ALL algorithms are moved to bob.pad.base package
    • Fixed algorithms documentation links, pointing to bob.pad.base now.
    • The documentation of the package now linked to joint documentation properly.
    • Refactoring: duplicated database configs are removed
    • Refactoring: removed VideoQualityMeasure, replaced with Wrapper class
    • Refactoring: removed VideoLBPHistogram, replaced with Wrapper class
    • Refactoring: removed unused VideoDataLoader
    • Refactoring: created FaceCropAlign, replaces ImageFaceCrop + VideFaceCrop using Wrapper
    • Added Spatially Enhanced Histogram option to LBPHistogram class
    • Changes to the FrameDifference preprocessor: Estimate the bounding box from annotations if not available. This is usually estimated from eye annotations. Remove the unused check_face_size flag. Fixed the conda recipe !63
  • v1.0.0
  • v0.0.1