Most of the video analytics solutions available on the market today are able to work on board of cameras (mainly on board of new generation cameras, equipped with accelerators) or on large servers (high cost, high energy consumption and typically equipped with GPU). What the market expects today is instead an embedded solution, low cost, low energy consumption, small in size but highly flexible. This solution, now available in the A.I. Tech portfolio products, is AI-APPLIANCE-DEEP.
It is an integrated hardware and software video analytics solution. It is based on the NVIDIA Jetson platform and in 75x88x39 cm it allows to run 4 different video analysis applications at the same time, choosing from over 20 A.I. Tech apps. Some of our customers have called this “best of breed artificial intelligence technology in a hand”.
The main advantage of this solution is its versatility: in fact, both old and new generation cameras can become “intelligent” thanks to AI-APPLIANCE-DEEP, without the need to change the cameras by choosing a “smart” one, capable of hosting third-party apps, or add a large and expensive server.
The solution is plug and play, accessible via an intuitive web interface. More than 20 different video analytics apps are available for managing any type of need for different vertical markets: smart surveillance, retail, business intelligence, smart transportation, smart road, smart parking, smart hospital, smart environments, smart city and many others!
All the apps can be dynamically activated, both time basis (on an hourly basis, for example from 9:00 to 20:00 every day) and event basis, depending on some external events occurring and activating the app dynamically via software, namely via HTTP requests.
Below, we briefly summarize the different apps available for some of the different vertical markets.
SMART TRANSPORTATION: responsible for monitoring vehicle flows (AI-ROAD3D) and for detecting abnormal driving behaviors (AI-INCIDENT). In particular, it is possible to count and classify the vehicles, and for each one to estimate their average speed and color. Furthermore, it is possible to detect vehicles stopped in prohibited areas, vehicles traveling in the wrong direction or on reserved lanes; also check for the presence of pedestrians along the carriageway or stationary in the lay-bys. Finally, it is possible to detect vehicle license plates, thanks to the combination of the most advanced deep learning techniques with a semantic engine that deals with automatic error correction (AI-LPR).
SMART PARKING for intelligent parking management, allowing to monitor the occupancy status of each space, providing information relating to the occupancy status (free or occupied) in real time (AI-PARKING).
SMART PEDESTRIANS for the control and monitoring of pedestrian flow, the management of crowds, and more recently the problems related to the spread of the Coronavirus infection that require careful management of gatherings and social distancing (AI-CROWD-DEEP), as well the control of people who do not wear the mask (AI-FACEDETECT). The system, therefore, offers support by counting people and estimating the number of people present within an area or the percentage of occupation, generating an alarm in case of overcrowding (AI-RETAIL-DEEP).
SMART ENVIRONMENTS through the early detection of the principles of flames (AI-FIRE-DEEP) and smoke (AI-SMOKE-DEEP), especially in environments such as urban ones where traditional fire detectors are not very effective or cannot be used. Monitoring the environment also means being able to detect in real time the presence of abandoned waste on the roadside, through the detection of abandoned objects (AI-LOST).
SMART SURVEILLANCE: the apps allow you to detect the presence of intruders, in terms of detection of permanence in a sterile area, line crossing, or multiple virtual line crossing (AI-INTRUSION), or even loitering (AI-LOITERING) with the possibility of distinguishing animals, people and vehicles thanks to a deep neural network.