Clothes Recognition

Functionality able to analyze in real time the faces of the subjects filmed by the associated camera, comparing them with a database owned by the customer, restricted and predefined, called "watch-list", whose size can be in the hundreds of thousands of subjects. The face recognition node sends an alert to the platform, that is to the command and control system, when a correspondence is found between the biometric model of the face imaged and the biometric model of the watch-list.
The system operator receives the alert generated following the comparison match (for example for an attentive subject who cannot access a place deemed sensitive or forbidden to it).

The biometric recognition baseline allows:

  • the acquisition of the biometric sample
  • the extraction of biometric traits and the generation of the biometric model
  • comparison with a reference whatchlist
  • to send customized notifications

Benefits:

  • real time analysis of subjects in a non-cooperative context
  • offline analysis of faces in previously captured video streams
  • comparison watch-list from any desired source
  • identification of all faces in the scene at the same time
  • face recognition in different poses (even extreme ones)
  • minimum distance between the eyes for the identification of a subject equal to 20 pixels
  • search in less than 50 milliseconds
  • real time alarm

Implementation

The main application areas are listed below:

  1. Public and private security management
  2. Access control management
  3. E-commerce
  4. Authentication and online payments

Background

" Private Companies / Public Bodies / State Police and Government"

To allow biometric recognition it's necessary to acquire the "biometric feature" (face) and from this extract the "biometric model" with a procedure that guarantees the correctness of accreditation in the biometric system (biometric enrollment), the link with the subject who subjects to the enrollment and the quality of the resulting "biometric model".

The biometric verification process is described below:

  1. The system extracts the biometric model from an identity document or a badge or image provided by the person;
  2. The subject declares his identity by lending his face to a camera;
  3. The system performs a compatibility check between the biometric model detected by the camera and the one previously acquired from the photo of the identity document and corresponding to the declared identity.

Implementation

The main application areas are listed below:

  1. Public and private security management
  2. Access control management
  3. E-commerce
  4. Authentication and online payments"

Background

Private Companies / Public Bodies / State Police and Government

The automatic license plate recognition algorithm (ALPR) is able to identify the number plates of vehicles entering or passing through a video surveillance area.

It can be integrated into a system that allows you to:

  • detect the license plate of a vehicle;
  • authorize access to a reserved area;
  • send a notification in real-time if the detected plate is present in a black list;
  • proceed with the necessary follow-up operations.

The operator of the command and control system can also search for a vehicle starting from the following characteristics: license plate; type (car, truck, motorcycle, etc.); template; color; time range (Date from - Date to; From Hours - To Hours).
The solution is plug & play and easily integrates with the customer's pre-existing infrastructure.

Implementation

It's possible to manage different scenarios of interest, by monitoring the road sections of the City and Highways. The system is usually used by the Police Forces to detect stolen vehicles or vehicles under administrative detention. The detected plates are compared with those of the vehicles reported.

Background

The system of detection and reading of the plates was successfully used for monitoring and security of the road arteries present in various countries

Emotion Recognition

The algorithm is capable of recognizing emotions and can be installed on any device equipped with an HTML 5 browser, without the need for an app or plug-in.

The algorithm is able to detect and report in real-time the presence of objects within a specific video surveillance area. Similarly, it allows the detection of an object removed from a video surveillance area.

The algorithm can be integrated into a system that allows you to:

  • detect the presence of the object by providing information about its position and size in an automated manner and in real-time;
  • classify the object;
  • send a notification to report the presence/absence of the object;
  • proceed with the necessary follow-up operations.

The solution is plug & play and easily integrates with the pre-existing infrastructure.

In the Public Administration area, the algorithm can be integrated into a system that allows the citizen to take a photo of the abandoned object detected in a specific place and send a report to the platform. The system analyzes all the acquired images offline and classifies the abandoned objects starting from a previously defined priority list. The Reco systems are able to recognize the following abandoned waste, classified into macro-categories: Building materials (Bricks); Abandoned vehicles (car, bicycle, scooter, etc.); Bulky items (shopping cart, sofa, chair, bench, etc.); Household appliances (washing machine, TV, refrigerator, etc.); Bathroom (Faucet, Sink, Toilet, etc.); Tires; Sacks.

Benefits:

  • detect abandoned objects avoiding the risk of accidents;
  • prevent the formation of illegal landfills and fires;
  • detect unattended objects and prevent the risk of theft or vandalism.

Implementation

  •  Monitoring public/private spaces as a control and security measure
  • Monitoring and control of road sections in the city and highways
  • Workplaces as a prevention and safety measure
  • E-commerce and User Generated Content solutions

Reco solutions are resilient to impersonation (spoofing) attacks, where the impostor directly uses a user's biometric data to attack or to create spoofs or fakes.
The RECO biometric systems based on Face Recognition have been successfully tested against the most common methods of Presentation Attacks and against Attacks using photos/videos:

  • printed photo attack (photographic paper or on A3 / A4 sheets);
  • warped photo attack (both vertical and horizontal);
  • photo display attack (playback on screen).

In addition, they have been successfully tested against attacks by video playback (either on smartphone displays; tablets or on laptop screens).

Applications

Prevention and security measure in the Banking / Insurance / TELCO / Private Companies / Public Bodies sectors

The algorithm is able to detect, track and interpret human movements.
The solution, created to analyze textual data and images, finds its main application in the recognition of human behavior through the identification and interpretation of the individual's movements.

The main application areas are listed below:

  • Detection of facial expressions and movements, of one or more subjects, considered anomalous or suspicious
  • Detection of a fast-moving crowd
  • Detection of one or more subjects who enter prohibited areas
  • Detection of one or more subjects attempting a theft
  • Detection of anomalous behaviors during the exam
    During the examination session, the system calculates in real time some parameters of the ICAO standard that allow you to identify abnormal behavior. In the specific case of an exam, the main anomalous behaviors are: absence of the examiner; examining, he diverts attention from the screen by turning his head for a time longer than a set parameter (in an attempt to "copy"); the presence of several people within the range of the webcam (someone is helping); the presence of a person other than the person being examined (someone else performs the task in place of the person being examined). The alerts of each candidate are monitored and the reports are sent to the person in charge of the examination session.
  • Monitoring of the attention level and facial expressions of participants in training courses (face-to-face and remote), video conferences and live events.

Implementation

  • Public and private security management
  • Training and Exams
  • E-commerce

The algorithm, in the experimental study phase, allows to identify the clothing items present in the image and to divide the element into sections.
It also allows you to classify and assign attributes. The main application areas are listed below:

  1. Visual recommendation systems: the system recognizes, through the presence of one or more rooms, the garment(s) worn by the customer. Starting from this information and possibly from the historical data associated with the user profile, it proposes a personalized list of products that could be of interest to you. The history of user-product interactions is completed with the visual characteristics extracted from the images of the products by means of machine learning models to generate the final recommendations. The customer views this personalized list and can provide feedback on each specific recommended product. The personalized list will contain products of the customer's liking and present in physical or digital stores.

  2. Generative models for the automatic and personalized creation of new garments: the system recognizes, through the presence of one or more rooms, the garment(s) worn by the customer. Starting from this information and possibly from the historical data associated with the user profile, it proposes new images, created ad hoc, which reflect the visual aspects that the model has learned. The customer views the images created.

  3. Augmented reality: the customer can select an item of clothing and / or accessory of his interest from a totem and perform a ""virtual test"" in real time. The segmentation mechanisms of the body (or part of it) will be used to which the images of the garment / accessory will be superimposed.

Implementation

  1. E-commerce
  2. Private companies for the improvement of the customer's customer experience