We are releasing brand new face tracking technology - check it out !

visage|SDK™ VISION package contains powerful functions to find and track faces and facial features in still images and video sequences in form of a well-documented C++ Software Development Kit.

Facial feature detection

The feature detector identifies the facial features in still images containing a human face. The input is an image containing a human face. Most standard image formats such as JPEG, GIF etc. are supported. The result is a full set of facial features such as the ones shown by the dots in this image:

The average detection time is 1.1 seconds measured on an Intel Core Duo processor (T5500) running at 1.66Ghz. The minimal size of the input image is 150x180 pixels with 8 bits per pixel (grayscale image; color images are internally converted to grayscale)

visage|SDK™ contains a sample project with full source code demonstrating the usage of the VISION package to find facial features in images.

Facial feature tracking (facial motion capture)

The feature tracker follows facial features in video sequences in real time. Input can be a video file or live camera image, in either case showing a human face. No markers or makeup are needed on the face. The captured motion can be used on-the-fly to animate virtual characters, or it can be saved in standard MPEG-4 FBA files. Furthermore, a mechanism is provided to allow developers to access the facial feature coordinates as they are being captured and use them in their own functions.

The minimal size of the face in the video image is approx. 80 x 120 pixels, meaning that the minimal resolution of the video image is approx. 100 x 150. Larger image gives better tracking results but speed is reduce so the optimum video resolution is between 320x240 and 640x480 pixels.

Currently the tracking algorithm supports only front-facing posture.

The tracking algoritm is person-specific and needs to be trained for a specific person before it can accurately track that person's face. visage|SDK™ contains an interactive training tool and instructions for performing the training. The training is based on a minimum of 30 images of the face with varying expressions and lighting conditions. For training it is better if the images have high resolution.

visage|SDK™ contains a sample project with full source code demonstrating the usage of the VISION package to track facial features in video coming from a file or from camera and to use the tracking results to drive character animation on-the-fly.

Next-generation facial feature tracking is coming!

visage|SDK™ VISION package is about to be upgraded with a brand new facial feature tracking (facial motion capture)  technology. For a preview of the new technology watch this video:

The new tracking technology supports out-of-plane head rotations and has a very simple setup procedure.

The new tracker is about to be released in the next release of visage|SDK- please contact us for more information or to request an evaluation.

Try visage|SDK now

Back to visage|SDK main page

CONTACT US FOR FURTHER INFORMATION