NEXYAD GIVING COURSES ON DEEP LEARNING AT MASTER LEVEL FOR STUDENTS OF INSA ROUEN (NORMANDY)
NEXYAD is part of the MOVEO Groupement ADAS that built a pedagocic chair with INSA ROUEN on Intelligent vehicles (autonomous Driving, ADAS, connected car).
NEXYAD presented neural netwoks theory, practice, and methodology to students of MATHS option (Génie Mathématique) – Mentor of the chair is Pr Aziz BENSRHAIR.
With examples of real time detection for intelligent cars :
. ObstaNex : obstacle detection
. RoadNex : road fre space detection by semantic segmentation (use cas of real world old dirty roads)
The methodology “AGENDA” (Approche Générale des Etudes Neuronales pour le Développement d’Applications) was presented to help future engineer in conception and maintenance of hybrid solutions involving deep Learning among other methods. Key questions are explained and solutions proposed : how to VALIDATE such a solution ? for instance.
But with no distraction, no anger, no rush, no mood, just like the day of patent when one have it… An Autonomous Driving System (AD System) is NOT a system that can detect “everything” and can react then the right way to “every detected situation”.
Human driver is much safer, in average, at 50 years old than at 18, although detection capabilities and reflexes are much better at 18. Improvement of safety comes from anticipation capabilities and application of safety rules in absence of detection.
NEXYAD proposes the technology SafetyNex that computes in real time at each moment the risk that the driver (human being or AD system) is taking. 75% of the Driving risk comes from the inadequation of Driving behaviour to infrastructure characteristics (so for those 75% no need for mobiles detection, and SafetyNex can assess this risk with only a GPS, a digital map, and accelerometers. With SafetyNex, the AD System is “AWARE” of the Driving risk it is currently taking and then can Apply heuristics of cautiousness even without (before) detecting any problem : i.e. slightly slowing down before a priority to the right … is a rule of cautiousness that avoids MANY accident without need for accurate ADAS detection.Driving behaviour becomes then more “human”
Here is a video of ObstaNex, by NEXYAD, detecting cars with a regular smartphone, feeding the real time Driving risk computing SafetyNex with additional inputs such as Interdistance, for instance.
ENHANCED safety functionality for everyone (every smartphone or modern dashcam owner). Detection results may vary a little depending on quality of camera, but anyway, it brings safety alerts and can save your life.
Already integrated into dashcams for intelligent telematics.
Gérard YAHIAOUI presented Artificial Intelligence SafetyNex at Meetup “Connected Vehicles : Technologies and Applications organized by Laurent Dunys, COO of Xmotion.
This meetup took place at Dunasys with a speech of Frédéric Lassara to explain Dunasys activities.
Thanks to “La Sécurité Routière” the french Road Safety for its invitation at Vivatech 2019 in Paris ! For the first time this governmental institution presented the innovation award winners on a common booth dedicated to safer roads. NEXYAD is very much involved with Road Safety. Since 2000, we have worked on many national and international research programs with experts of 19 countries included Europeans, US and Japanese. Receive an award for SafetyNex is a great accomplishment for us and for our integrator customers.
We made a lots of precious contacts during this amazing show.