Tevian Software Development Kits

Real-time face recognition and processing

Cross-platform C++ library that includes all algorithmic functionality for analyzing human face images. Provides maximum efficiency and flexible integration into third-party software. Well suited for integration into video surveillance systems (Video Management System), access control and management systems (ACMS), self-service terminals, mobile applications.

TOP-1 in speed and quality at ECBI

Capabilities

• ●
  Face detection in images
• ●
  Face tracking in video, best frame selection
• ●
  Face image quality assessment (overall, rotation angles, blur, occlusions)
• ●
  Biometric template creation and comparison: verification (1:1 comparison), identification (1:N comparison)
• ●
  Demographic classification by face (gender, age)
• ●
  Liveness detection
• ●
  Face mask detection
• ●
  Attribute classification: presence and type of headwear, glasses, mustache, beard, etc.
• ●
  Additional attribute classification: presence and type of headwear, glasses, mustache, beard, etc.
• ●
  Gaze direction determination
• ●
  Phone conversation detection
• ●
  Smoking detection
• ●
  Deepfake detection

Vehicle recognition in video. Determines type, color, make, license plate

Cross-platform C++ library implementing vehicle image analysis functions. Provides maximum efficiency and flexible integration into third-party software. Well suited for integration into video surveillance systems (Video Management System) and access control and management systems (ACMS).

Capabilities

• ●
  License plate recognition in images and video
• ●
  Number recognition with readability assessment and classification by type and country
• ●
  Vehicle silhouette detection and tracking in video
• ●
  Vehicle category determination
• ●
  Special vehicle detection
• ●
  Vehicle make, model and color determination
• ●
  Driver seatbelt detection
• ●
  Mobile phone usage detection while driving

Human detection and tracking in video, attribute classification, re-identification

Cross-platform C++ library designed for human detection in images. Also includes web service for web API integration.

Capabilities

• ●
  Human silhouette detection and tracking in video stream
• ●
  Re-identification – person search by sample based on shape, silhouette, clothing, etc.
• ●
  Multi-camera tracking
• ●
  Heat map generation
• ●
  Attribute classification: clothing, headwear, bags, etc.
• ●
  Fall detection
• ●
  Smoking detection

Recognition of passports, driver's licenses and other standard documents

Cross-platform C++ library implementing standard document recognition functions. Provides maximum efficiency and flexible integration into third-party software. Well suited for integration into access control and management systems (ACMS), self-service terminals. Also includes web service for web API integration.

Capabilities

• ●
  Russian Federation passport
• ●
  CIS ID card
• ●
  Russian SNILS
• ●
  Russian driver's license
• ●
  Vehicle registration certificate (VRC)

Assessment and analysis of people presence in frame

Cross-platform C++ library and web service based on it for estimating number of people in frames. Allows counting people even at very low resolution (from several pixels).

Capabilities

• ●
  Venue attendance monitoring
• ●
  Room occupancy fluctuation analytics
• ●
  Public place overcrowding control
• ●
  People interest zone detection

Real-time operator state analytics and attention monitoring

Web service and C++ library for real-time monitoring of operator and employee state. Provides tools for detecting signs of fatigue, distraction, dangerous behavior and misconduct. Ensures maximum efficient integration into third-party software: video management systems (VMS), workforce management systems (WFM), HR systems and analytics platforms. Suitable for integration into contact centers, dispatch centers, security complexes, vehicle fleets and production facilities.

Capabilities

• ●
  Face detection in frame and tracking in video stream
• ●
  Fatigue detection: recognition of yawning, closed eyes, characteristic head movements
• ●
  Attention distraction detection: analysis of gaze direction and head trajectory
• ●
  Dangerous behavior classification: smoking, eating, drinking, mobile phone usage
• ●
  Personal protective equipment detection (helmet, vest, glasses)
• ●
  Statistics and reporting generation by time and violation types
• ●
  Integration with alerting systems for operational response

Railcar number recognition and railway rolling stock inventory

Tevian Railway SDK is a software module for integrating video analytics into railway logistics systems, industrial sites, ports, terminals, depots, and stations. The SDK recognizes numbers of freight and passenger railcars, locomotives, and flatcars, tracks objects in the video stream, and helps automate railway rolling stock inventory and monitoring. The solution is powered by computer vision algorithms and reliably recognizes markings in challenging operating conditions, including low light, adverse weather, body contamination, worn or partially damaged numbers, train movement, stops, and changes of direction. Railway SDK can be embedded into existing information systems, video surveillance infrastructure, rail weighing systems, and transport monitoring platforms.

Capabilities

• ●
  Railcar and locomotive number recognition. The SDK automatically reads 8-digit freight railcar numbers with checksum validation, two-line passenger railcar numbers, and non-standard locomotive number formats.
• ●
  Recognition in motion. The system operates reliably at train speeds from 2 to 50 km/h and handles stops, reverse movement, and direction changes during shunting operations.
• ●
  Performance in challenging conditions. Railway SDK recognizes markings in low light, adverse weather, body contamination, partial number damage, and wide-angle lens distortion.
• ●
  Multi-camera data aggregation. The SDK tracks railcars in the video stream and combines results from cameras installed at different viewpoints, including the left side, right side, and underside. This improves recognition reliability and reduces the risk of missed numbers.
• ●
  Identification without external sensors. The solution is based on video analytics and does not require magnetic loops or dedicated sensors to detect the beginning and end of a railcar.
• ●
  High recognition accuracy. The successful recognition rate exceeds 92% with a single camera and reaches 96–98% with two-sided video capture.