Core Technology

  • Face Recognition

  • Traffic Recognition

  • Other Technology

  • Face Detection

    Locate the faces in the photo precisely; the detection rate of this technology on the global face detection dataset FDDB is 98.74% reaching the world advanced level.

  • Facial Landmark

    Locate facial landmark precisely; have strong adaptability to special circumstances, such as wide-angle shot from the side, expression change, occlusion, and shading

  • Face Attributes

    Extract facial features based on convolutional neural network of deep learning. Analyze the features for achieving facial features recognition

  • Celebrity Recognition

    Identify the celebrity information and corresponding probability that best match the image

  • Traffic incident detection

    In the case of complex road condition information such as urban roads or expressways, the camera video can be used for traffic event early warning and positioning through roadside analysis equipment。

  • Traffic parameter statistics

    Through the analysis of the operating traffic parameters of all vehicles on the road, some reference and conclusive data display can be made for the traffic situation, which is convenient for the expressway management department to optimize the control, traffic guidance and vehicle road coordination。

  • Vehicle feature recognition

    Detect the position of the vehicle in a picture and identify the model and color of the corresponding vehicle. The identifiable models include sedan, van, pickup truck, off-road, bus, etc。

  • License plate recognition

    Detect the position of the license plate in a picture, recognize the license plate number, and return the corresponding license plate number。

  • Pedestrian detection

    The depth learning technology is used to judge whether there are pedestrians in the image or video sequence and locate them accurately at the same time。

  • Parts defect detection

    Carry out standard parameter comparison and judgment on the specified details of parts, so as to provide judgment and research standards for the configuration of automatic equipment in the quality inspection production line。

  • Component dimension measurement

    Through image correction, image enhancement, smooth denoising, binarization, morphological operation and edge detection operation。

  • Super resolution

    Using the reconstruction algorithm, the blurred image in the video or picture can be restored to an easily recognizable high-definition image。

  • Living discrimination

    Detect whether the user in front of the camera is a real person through the specified action analysis (shaking head, nodding, blinking, opening mouth, etc.)

  • Porn/terrorist recognition

    Enter the photo, and judge the photo to be pornographic, sexy, violent, and so on。

Products And Services

Video structured platform
  • Product Background

    With the development of Ping An City, video surveillance network sharing and more and more video materials, video structuring can make efficient use of these massive video information resources, solve some alarm monitoring problems in real time, quickly locate the human and vehicle information to be understood, and contribute to the risk assessment, event early warning and judgment of social security monitoring.

  • Product Function

    Video structuring refers to the establishment of a video big data structured platform according to the characteristics of people, vehicles, objects, colors, numbers and other attributes presented in the video picture. In particular, it provides a deeper structural analysis of the objectives of people, vehicles and non motor vehicles. After the video is structured, it is stored in the corresponding structured data warehouse.
    In the application layer, the video structured analysis server can find the suspects and vehicle targets in a screenshot from the video database for the structured system in the video image, and can study and judge the behavior process of the target through the monitoring data collected from different locations.

Industrial vision production line solution
  • Product Background

    The 14th five year plan and the long-term goal of 2035 clearly point out that we should promote the deep integration of Internet, big data and artificial intelligence, and promote the development of advanced manufacturing clusters。
    The characteristic of industrial machine vision system is to improve the flexibility and automation of production. Machine vision is often used to replace artificial vision in some dangerous working environments that are not suitable for manual operation or where artificial vision is difficult to meet the requirements; At the same time, in the process of mass industrial production, the efficiency of product quality inspection with artificial vision is low and the accuracy is not high. The machine vision inspection method can greatly improve the production efficiency and production automation。

  • Product Function

    According to the actual needs of industrial vision of manufacturing enterprises, flying search can provide a complete set of visual inspection equipment, including industrial camera and lens, light source, control unit, image processing computing equipment, supporting recognition algorithm and specific implementation scheme of the project。

Application Technology

According to the actual needs of enterprises, flying search can provide more production line solutions based on industrial vision, including recognition model training, hardware system deployment and production line assembly and debugging, and provide test data of various dimensions. The specific scenarios are: visual guidance and positioning, pattern recognition and detection, accurate measurement and ranging, product appearance detection, etc。

The system is actually implemented in a city in East China to complete the quality inspection defect detection in the production link and realize large-scale operation. In the practical application of the project, it greatly reduces the investment of enterprise quality inspection manpower and improves the quality inspection accuracy。

Computing Server
  • Product Description

    Faceall-1000 face static/dynamic comparison server mainly provides face detection, face tracking, feature extraction and retrieval in photos and video streams. It supports the management of face images of millions and tens of millions of scales, high-speed feature extraction, and rapid large-scale concurrent search. The accuracy of comparison Results is high, and the number of returned Results is controllable. Besides, the product is flexible to deploy which can be used as a stand-alone machine or be gather together for forming a powerful and flexible cluster.

  • Key characteristics

    high cost performance and scalability
    Face search speed in ten million level database is less than seconds
    recognize faces accurately in complex and variable situations
    Support more than 10 features recognition
    Rich and friendly interfaces for business development

Face detection and recognition
  • Product highlights

    Face recognition technology can quickly and accurately identify the location of the face in the picture, which can be used for face beauty map, interesting face change, automatic special person search in news video, designated suspect in public security inspection, marriage faces, fate and other fields.

  • Product description

    Face detection is to recognize the face in the image through the algorithm of deep learning and machine learning. The position information of face is given. Celebrity recognition can identify which celebrity the face in the figure is most like. The system will compare the selected face with the predefined 20000 celebrity faces and give the top 50 celebrities that are most similar to it. Compare the similarity of two faces to judge whether they are the same person, and the score is between 0-100%.

Object recognition
  • Product highlights

    The service can automatically identify the target categories and scenes in the image (thousands of categories are customized, including clothing, home, animals and plants, violence, pornography, etc.) and can be used in product recommendation, automatic album analysis and sorting, Internet Security and other fields.

  • Product description

    Using the deep learning method, quickly and accurately identify the category and scene of the target in the image, and the system will mark the 2000 categories (including pornographic image recognition and violent and terrorist image recognition) defined by us on the image. The probability output range is 0% ~ 100%. The higher the score, the closer the test picture is to the semantic concept. According to specific settings, up to 50 similar categories can be returned. Among them, pornographic identification is divided into four categories: pornographic, suspected pornographic, sexy and ordinary.

Offline SDK License
  • Product Description

    The product provides interfaces for different language environments, adapts to different datasets, and provides redeveloped according to different application scenarios. Faceall products have been widely used in security, finance, transportation, airports, railway stations, communications, hotels, education, electricity, justice, exhibition halls, etc.

  • Product Form

    Authorization method is combination between dynamic library and dongle (soft and hard)
    Supporting for Linux, Windows, dataset
    Support 10+ feature recognition
    Some functions support Android and IOS datasets

Product Function

The complete algorithm module required by the face recognition system, video structuring system and crowd analysis system

The face recognition system mainly includes face detection, facial landmark detection, face gesture, face tracking, face quality assessment, facial features analysis, facial features extraction, face 1:1 comparison, face 1 :N comparison, face M:N comparison and the supporting functions of specific scenes, such as the physical detection required by the bank to open an account, OCR recognition of ID card, and the facial super-resolution for the police.

Video structuring system mainly includes pedestrian & vehicle detection, pedestrian & vehicle tracking, pedestrian & vehicle features analysis, license plate location and identification, license plate super resolution and other functions.

The crowd analysis system mainly includes real-time crowd density map, population statistics, crowd flow, and crowd behavior analysis.

Online demo

Face Verification

Demonstration scope: Compare the similarity of two faces, the probability output range is 0%~100%. The higher the score, the higher the similarity between the two faces

Application scenario: payment authentication, face authentication in airport and customs, automatic photo classification

  • Local Upload

    Local Upload

  • Response Json
Face Attributes

Demonstration scope: (1) 68 facial landmark detection represented by eye, nose, mouth and overall contour. (2) features include age, gender, and facial expression (probability output 0%~100%).

Application scene: face retouch, face change for fun, marriage intermediary

  • Local Upload

  • Results Response Json
Celebrity Recognition

Demonstration scope: Identify the celebrities or similar celebrities in the photo through deep learning algorithms.

Application scenario: specific person search in news videos automatically, investigate the designated suspect.

  • Local Upload

  • Results Response Json
Porn Recognition

Demonstration scope: The system will mark pornographic and violent information in the images, and judge that whether the map is pornographic, sexy, suspected pornography, etc.

Application scenario: public security system, internet security

  • Local Upload

  • Results Response Json
Vehicle Recognition

Demonstration scope: The system will analyze the detailed structural information of the vehicle contained in the uploaded picture, including the color, type and brand of the vehicle, the color, type and number of the license plate, etc.

Application scenario: public security system, traffic control system.

  • Local Upload

  • Results Response Json

Company News