We’re delighted to announce the launch of our new CrowdSight SDK version 3.0. We increased the accuracy of existing analytics, made our SDK more robust and added several new features in order to meet our clients’ needs. We are now able to not only detect frontal faces, but also side faces; additionally, we now offer new functionality to have more control over face recognition (identify) and verification (compare), and to determine ethnicity of audiences.
CrowdSight SDK is a crowd face analysis software that detects multiple faces at once and retrieves relevant information like facial expressions, age, gender, mood, head pose, clothing style and from now on, ethnicity as well. Companies are able to obtain a better understanding of their audience while they behave spontaneously in their natural environments. This is possible by using a simple webcam or a regular external camera.
Features of the new version 3.0:
- Model identification and comparison: Crowdsight can now load, save, compare and identify face models.
- Ethnicity Classification: CrowdSight identifies the most probable ethnicity group of the detected faces (African, Asian, Caucasian or Hispanic).
- Adaptive Face Acceptance Threshold: the system learns how faces look in your current setup and uses that information to detect false positives.
- Side Face Detection: the SDK is now able to detect faces who do not look in the direction of your camera.
- Reset Function: reboots the memory for video batch processing, without requiring a reload of the classifiers.
- Updated Age Estimation: we significantly improved the accuracy of the age classifier by using an extended training set.
- Updated Gender Estimation: gender accuracy was also improved thanks to an extended training set.
- Updated Asian Mode: the Asian database has been improved with the help of an extended training set, hence significantly improving the accuracy of the age and gender classifiers.
- Asian mode settings can now be changed without restarting the program.
- Night mode settings can now be changed without restarting the program.
- Region of interest settings can now be changed without restarting the program. This allows to specify a region where the faces should be detected.
- New built-in Face Recognition example: this new example demonstrates how easy it is to build face models and compare them.
Please be aware that version 3.0 is not fully backward compatible with previous versions of CrowdSight, meaning that you will need to apply some minor modifications to your existing code in order to upgrade.
Upgrade your CrowdSight to the latest version today and get the advantage from all the new improvements and functionality for yourself.
Download our demo to get a glimpse of the new CrowdSight 3.0.