Technology

Visidon offers computationally efficient video analysis technology for several embedded and mobile device platforms in addition to personal computers.

• Face Detection and Tracking
• Face Identification
• Gender, Expression, and Age Recognition

Face Detection and Tracking

Visidon's face detection and tracking technology improves image capture and post-processing based on human faces. Furthermore, facial information enables novel and innovative applications.

State-of-the-art in accuracy

• Any orientation, including in plane and out of plane rotations
• Unlimited amount of faces per image
• Simultaneous tracking of faces ranging from whole image to 1/16 of image width in size
• Robust to varying imaging conditions and input image (illumination, sharpness, color, noise, etc.), allowing effortless image quality improvement

Fast detection and tracking

• Typical initial face locking time less than 100 ms, ARM9 180MHz, QVGA input
• Tracking 10 ms per face

Designed for embedded usage

• Windows, Linux, Windows Mobile, Symbian, ARM, x86
• Easy to port to new platforms
• Small code size and memory usage
• Possibility for customer specific tailoring and optimization services

Face Identification

A vast majority of images and videos contain people. With our recognition technology, these faces can now be identified.

The technology is highly optimized for embedded devices and is easy to integrate to digital cameras, mobile phones, surveillance cameras, and other imaging devices.

State-of-the-art recognition performance

• 70%-99% against 1000 individuals with number of standard and private databases
• Tolerates varying imaging conditions and illumination
• Robust to aging (including beard, glasses and hair style change) and expressions

Fast processing time

• Typically 1 second for the input face (ARM9,180MHz)
• Fast comparison (<1 ms per person in database)
• Adaptive for optimal application performance

Designed for embedded usage

• Windows, Linux, Windows Mobile, Symbian, ARM, x86
• Easy to port to new platforms
• Possibility for customer specific tailoring and R&D services

Gender, Expression, and Age Recognition

Face information, complemented with age, gender and expression recognition, enables creating novel and innovative applications.

State-of-the-art performance

• Over 90% accuracy for gender measurements
• Average precision of 6 years in age estimation
• Expression categorization accuracy of 90%

Fast processing time

• Real-time smile and blink detection on mobile phones
• Allow pre-capture analysis

Designed for embedded usage

• Windows, Linux, Windows Mobile, Symbian, ARM, x86
• Possibility for customer specific tailoring and R&D services