AI/ML 
Rozwój oprogramowania 
AI Video Analysis System for Real-Time Vehicle Detection and Counting
Client
The project was delivered for an organization seeking to implement a real-time video analysis system for monitoring and vehicle counting purposes.
Challenge
With the growing use of video surveillance systems in areas such as public safety, traffic analysis, and agricultural monitoring, the need emerged for a tool capable of real-time video stream analysis. The key requirements included high-precision object detection and tracking, support for multiple operation modes, handling of multiple video sources, and browser-based access to the service.
Solution
A real-time vehicle tracking application was designed and deployed — adaptable to other object types (people, animals, plants) and scalable across various business scenarios. The system is built on a microservices architecture, where each component runs in an isolated Docker container. It integrates a deep learning-based object detector and multi-object tracking system.
The application operates in three core modes:
- Single image detection via REST API with ultra-low latency (~450 ms),
- Peer-to-peer streaming with end-to-end latency of approximately 2 seconds,
- Client-server streaming architecture, optimized for a wide range of user devices.
The system supports multiple video sources (IP cameras, webcams, video files) and allows multiple clients to simultaneously access an annotated stream. Bidirectional communication enables real-time analysis and seamless result delivery directly to the user’s browser.
Results
The system enabled:
- Counting vehicles moving in any user-defined direction,
- differentiation between vehicle types (cars, trucks, motorcycles),
- reliable performance in both local and internet-based streaming environments,
- fast adaptation to other use cases (e.g., detection of people or animals), making it a versatile tool for multiple industries.
Technology
The system was developed using:
- Frontend: React.js, TypeScript, Redux, Redux-Saga, Styled-Components
- Backend: Python, PyTorch, OpenCV, AsyncIO, Multiprocessing
- Video streaming: WebRTC and HLS
- Infrastructure: Microservices deployed in Docker containers
Jak usprawniliśmy rozwój i utrzymanie systemu ERP w aplikacji typu legacy
Design, Development, DevOps czy Cloud – jakiego zespołu potrzebujesz, aby przyspieszyć pracę nad swoimi projektami? Porozmawiaj o swoich potrzebach z naszymi specjalistami.
Jakub Orczyk
Członek zarządu / Dyrektor sprzedaży
Zamów bezpłatną konsultację