CrowdSight Changelog

v3.9.0

• Update Asian Age and Gender classifiers
• Improved detection of machine ID for Linux. More Linux devices are now supported

v3.8.1

• Expose getMachineID() to retrieve the machine ID that gets registered on the Sightcorp license server.

v3.8.0

• Ethnicity Asian bias removed
• Support GCC5 on Ubuntu/Linux
• Adding “runtime directory” feature. The runtime directory can be specified in the CrowdSight constructor and will be used for writing settings.ini and auth_key files, while the data directory can remain read-only. If no runtime directory is specified, the data directory will be used as runtime directory as well.
• Fixing bug related to static memory deinitialization

v3.7.0

• Introduce EvidenceLevel in settings.ini: this parameter allows for quicker ID assignment on observations if the level is lowered, or builds up more evidence for ID assignment when the level is increased. Default level=3. This parameters indicates the number of detection required before assigning an ID to  the user.
• More robust Age/Gender classifiers
• More robust Asian Age/Gender classifiers
• VS2015 support

v3.6.3

• More robust Asian Gender classifier
• getNumThreads and setNumThreads to control the number of threads CrowdSight can utilize
• Bugfixes

v3.6.2

• firstSeen for Person interface
• Speed optimizations
• Bugfixes

v3.6.1

• Android SDK release with fully updated interface
• TBB integration for Android
• Upgrade OpenCV for Android to 2.4.10

v3.6.0

• TBB integration for desktop: CrowdSight can speed up processing time by utilizing multiple threads
• Complete set of getter/setter functions for runtime adjustable CrowdSight settings
• Improved Face Tracking
• Improved Age/Gender classifiers
• Upgrade OpenCV to 2.4.10

v3.5.2

• Speed optimizations
• Bugfixes

v3.5.1

• Bugfixes

v3.5

• Landmarker points of the face
• Illumination normalization
• Headgaze feature
• Improved Ethnicity classifier
• Improved Age/Gender classifiers
• Further decreasing size of classifiers to decrease overall package size

v3.0

• Introduce Ethnicity classifier
• Improved Age/Gender classifiers
• Face Identification module
• Face tracking module
• Decrease size of classifiers to decrease overall package size

v2.2

• Mobile version
• Adaptive face acceptance thresholding
• More consistent model load and save
• Changes to the settings file
• Changes to the "person" interface for interchangeability with insight
• Averaging buffer is now internal to the SDK
• New example project, uses make for compilation on Linux and Mac
• Bug fixes

v2.1

• Night vision mode for applications that use night vision camera
• Asian mode for improved accuracy on asian countries

v2.0

• New function getEmotions, returning responses for the 6 universal emotions
  
• Adaptive filter becomes more flexible for training with new adjustable settings 
• Improved gender classification
• Integration new licensing server and developer/redistribution licensing scheme
• New function setMaxNumPeople to regulate the maximum amount of people to be tracked simultaneously.