The Jacaré Face Tracking is our facial recognition software. It analyses the stream of images provided by the camera. It provides instantaneous and continuous key information about people in the field :
Face detection : instantly detects all faces in the scene, attention time, position and distance.
Gender and age classification of each face detected.
Scene analysis in real time : how many persons are passing by the scene, unique person counting and counting by gender.
Face Tracking & Identification : tracks faces as long as they remain in scene and short-time person ID recognition for unique person counting.
High accuracy : The accuracy of gender recognition reaches 90% and that of the age estimator is below 8 years of difference. This performance is guaranteed in several situations and over various distances.
Intelligent counting : Our system uses artificial intelligence to anonymously identify people and to assign a unique identity number.
Simplicity : Our setup is quick and easy for large-scale deployments.
Speed and Flexibility : It is a multi-platform solution (Windows / Linux / Raspbian) and works with most cameras.
Privacy : No image is ever recorded and no uniquely identifiable data are extracted.
Please note that a new license is required for each installation.
The license type is a unique activation key per machine. It can be delivered via e-mail.
For more information, please contact Jacaré Technologies support on 01 77 93 21 95.
The Jacaré Face Tracking is a facial recognition software that collects and analyses demographic information in real time from video streams. The application supports several types of videos and streams captured by cameras. It collects information about gender, age and identity of the people who are in the field.
Jacaré Face Tracking’s functionalities comprise facial detection, recognition of gender and age, people identification and analysis of the situation in real time.
• Facial Detection : The application detects all (front) faces that are in the field. These data are then transferred to other components for further analysis. Our facial detection combines several advanced techniques in the field of object detection to eliminate false positives while maintaining a real-time processing speed.
• Gender and Age Recognition : Our module uses each image received from the Face Detector to give an instant estimate of each face. Gender and age recognition results are therefore available immediately as soon as a face is detected; they are also aggregated in time to become more reliable and stable if the person stays longer in the scene. Our module uses state-of-the-art recognition technologies that allow it to maintain a very honourable and stable performance, even on very low resolution (25 × 25 pixels) images.
• Person ID : We assign a unique identity number to each person detected in the scene. This identity is kept during one hour to trace this person through his/her multiple passages and to give as reliably as possible an estimate of his presence time and his attention time.
• Real-time situation analysis : Jacaré Face Tracking application provides a summary of the situation in real time, such as the number of men / women and the counting of unique people. This information is very useful in several situations, for example for advertising targeting. All these features are delivered to customers via a carefully defined JSON scheme so that they can easily integrate the Jacaré Face Tracking into their products.
Potential applications :
• Advertising targeting
• Measure of audience and Analytics
• Facial Recognition
• Person counting
• Human Machine Interface
• Augmented reality
Face Tracking software uses FFMPEG library to decode and play videos and video streams from cameras. It supports most common codecs and video formats. Clients can find a list of video codecs supported by FFMPEG at http://www.ffmpeg.org/general.html#Video-Codecs.
Jacaré Face Tracking can handle video up to 1280 x 720p (*)(**). The software is not intended to process each individual image.
(*) Tested on a NUC i3 6100U 2.30 GHz installed with Windows 10.
(**) Videos with higher resolution is resized before processing.
Our facial recognition was tested on:
• Windows 8/10 64-bit
• Ubuntu / Debian 64-bit
• Raspbian of Raspberry Pi 3
Other platforms may be considered upon request. In this case please contact us directly.
(1)* offer subject to conditions