The backbone of computer vision.

OVERVIEW


we use faces to relate to the world

FEATURES


face detection in a nutshell

INTEGRATION


and performance specifics

OVERVIEW

blue_small_bar

we use faces to relate to the world

FEATURES

blue_small_bar

face detection in a nutshell

INTEGRATION

blue_small_bar

and performance specifics

OVERVIEW

  • NAME: FaceDetection & Tracking
  • USAGE: detect and track faces
  • HOW: it uses analysis and registration techniques
  • IP: FaceDetection is a patented technology
  • PART OF: FacePower product family
  • TYPE: computer vision product, user experience technology
  • INTEGRATION: software and hardware (via IPU)
  • MARKETS:    1 2 3 4  7  5 8 9 10 11 12

Simply put, FaceDetection is the backbone of computer vision.

FotoNation is the first company in the industry to successfully bring to market and secure mass adoption of face detection technology. FaceDetection has enabled many consumer camera features and is a key component of enhancing very-fast-focus times.

From the first FaceDetection solutions implemented in digital cameras starting over a decade ago, to the best smart devices in the world today, FotoNation has continuously strengthened the performance of our face detecting technology. Our foundational technology enables the highest-possible performance when taking pictures of people and faces.

FotoNation’s FaceDetection is the core component of some of today’s most compelling imaging technologies, such as the world’s fastest auto focus and world’s best face beautification technologies.

FEATURES

80% of the pictures taken today contain at least one face. As a main enabler for core technologies that can help us document our memories, FaceDetection has become a key component in products that beautify the face, add make-up, focus faster and more accurately, recognize you, or understand your emotions to make decisions based on it.

Our FaceDetection & Tracking technology has been in a process of continuous improvement for over a decade. This allowed us to achieve outstanding performance metrics such as 14 millisecond locking time and distance detection of almost 8 meters. This is possible due to the hardware implementation of our solution and our unique hybrid software – hardware implementation that assures the best mix of performance, power consumption and flexibility while being a scalable, cost-effective and future-proof solution.

Excellent general-purpose detection in extreme illumination conditions and face poses.

No practical restrictions on the number of faces that can be detected in a wide range of poses. Scene-dependent parameters are collected and used to improve accuracy.

Implements advanced tracking algorithms, so it is not reliant on template matching alone. Once the face is “locked” (detected for the first time in the video stream), it will still be tracked, even if its pose changes dramatically.

Uses full color information which allows for a very low false positive ratio, but it can also work on grey-scale images.

Self-calibrating to difficult lighting conditions, helping to optimize and improve the quality of the image and movies (back-light illumination conditions, under/over exposure, low lighting conditions).

With orientation sensor availability, face tracking can provide face orientation information (0, 30, 90, 120, -30, and -90, -120 degrees of in-plane orientation, and semi-profile and profile (left or right) for out-of plane orientation).

INTEGRATION

FotoNation Face Detection and Tracking is available as software only or with hardware acceleration, taking advantage of FotoNation IPU cores.

The hardware engines enable best-in-class performance capability at a fraction of power consumption compared to conventional solutions.

The software library can be adapted during integration to provide optimum performance for a large number of applications and usage models.

The library is optimized for embedded use: integer-only algorithm, inbuilt memory management system for high performance, easy integration and porting, efficient run-time memory footprint requirements and a flexible API.