Affdex SDK for macOS and iOS

Updated the face detection to fix known bugs and improve its robustness for detecting non-frontal faces
Updated the brow raise, brow furrow, cheek raise, nose wrinkle, upper lip raise, mouth open, and smile models
Updated the emotion models, valence, and expressiveness to use the new versions of the expression models
Updated the gender model for improved accuracy
Added the TensorFlow runtime to be able to run deep learning based models.
Added support for Carthage

The detector has been updated to use the new DE Pack. The new pack improves
classifier accuracy for both emotions and expressions.
The AFDXFace interface now contains a new faceQuality member of type AFDXFaceQuality,
which includes a property named "brightness". This field indicates how well the face
is lit for purposes of analysis. The value range is [0, 100]. There are no hard
boundaries, but the following ranges can be considered as general guidelines:
0-10: too dark
10-25: suboptimal
25-75: well lit
75-90: suboptimal
90-100: too bright
The deprecated property face.expressivenessScore has been fixed to return non-zero
values. Note that this property is deprecated and that face.emotions.engagement
should be used instead.
For macOS only, a new initWithDelegate variation supports horizontal mirroring of
images before they are processed by the detector. This allows mac apps to choose
whether the display will show as right-reading or mirrored. For iOS, the mirror
setting is determined by whether the front-facing camera is in use.
For macOS only, a frame latency issue has been fixed. Previously, slower frame
rates were seen for frames where no face was detected. With this change, the
frame rate will be the same or slightly higher than the face-found case.

Removed the requirement for a license string to use the SDK. The AFDXDetector
properties licenseString and licensePath are deprecated, and any app code which
sets these properties can now be removed from apps which use the SDK.
The SDK is now packaged as a CocoaPod instead of a downloadable framework.
[iOS only] Added support for logging of firehose analytics records through AWS.
See the new AFDXDetector methods enableAnalytics and disableAnalytics in Affdex.h.
Added support for age and ethnicity classifiers.
Added new appearance classifiers: cheekRaise, dimpler, eyeWiden,
jawDrop, lidTighten, and lipStretch.
Fixed an out-of-memory error when trying to allocate memory for an image.

Addressed an issue where didStopDetectingFace: did not pass the AFDXFace object properly
The Device_Framework no longer contains simulator architectures