InSight Changelog

v3.2

  • Mobile version
  • Significantly improved speed on face tracking
  • Speed improvement on getAge and getGender
  • added "person" interface for interoperability with CrowdSight
  • GetAge and GetGender do not require  a face detection anymore
  • New example project, uses make for compilation on linux and mac
  • Bug fixes

v3.1

 

  • Major improvements to gaze estimation:
 
We have enhanced our gaze estimation algorithms to be more invariant to changes in head pose.
Further, we have added a default calibration, which allows you to start working with gaze estimation without necessarily requiring to calibrate every time.

  • Free contextual gaze correction:
 
Our contextual gaze correction function no longer uses credits. Improve your gaze estimation results for free.

  • Estimate gaze calibration quality without use of credits:
 
We have added a function that allows you to estimate the error of your gaze calibration. This way you can re-calibrate if you are not satisfied with the result, without using any credits.
 
  • Estimate the quality of the tracking:
 
We have added a function that allows you to estimate the quality of the mask that is fit to the users face during tracking. You can now maintain your own quality standards in order to achieve even better results.

  • Free contextual gaze correction:
 
Our contextual gaze correction function no longer uses credits. Improve your gaze estimation results for free.
 
  • Age and gender classification on demand:
 
Age and gender classification can now be turned on or off on demand. When not using these features of the SDK, the size of InSight package can be reduced by 60%. Shipping only essential parts of InSight saves bandwidth and waiting time for you and your end-users.
 
  • Use of external face detector:
 
InSight can now be used with your own face detection algorithm. You can specify where the face is in the image and inSight will pick it up and process it. This allows to select a face that you want to process in images that contain multiple people.



Have more questions? Submit a request

0 Comments

Article is closed for comments.
Powered by Zendesk