20.04.26 (inclusive)

Monetary reward (payment by stages / sprints).

MVP

ICT tasks

Car industry

Intelligent control systems

Software/ IS,

Mobile app

Architecture and Accessibility (Lock Protection) To ensure 100% fault tolerance and independence from messengers, the development is based on the Hybrid Web/Mobile model.: • Single Backend: Centralized processing of all AI logic and databases. • Frontend: Progressive Web Application (PWA) with mirroring in Telegram Mini App. The service must be available both via a direct link in the browser (with the installation of an icon on the desktop) and inside the messenger. 3. Key modules and functionality Module 1: Video Analytics and AI (Computer Vision) • Detection: Using YOLOv8/v10 models to detect cars in 6+ zones simultaneously. • Recognition (LPR): Integration of libraries (Nomeroff Net or analogues) for stable reading of RC numbers in any light. • Activity analysis: Detection of stages (start/process/finish) based on visual data (analysis of movement and changes in ROI). • Streaming: Stable RTSP video capture with automatic stream recovery in case of interruptions. Module 2: Predictive Core (ETD Engine) • Zero-Manual Calculation: Automatic calculation of departure time (Estimated Time of Departure) based on: 1. The selected type of service (Standard /Express). 2. The current stage of the process, visually confirmed through AI. 3. Historical data on the average service time at a particular facility. Module 3: Queue Manager (Logic Backend) • Dynamic Rescheduling: Cascading recalculation of schedules in case of deviations (early departure or delay). • Notifications: Automatic mailing (Web Push/ Telegram) about queue shifts. • History: Data storage (car number, session time, check-in/check-out photo). 4. Technology stack • Backend: Python 3.10+ (FastAPI), PostgreSQL, Redis. • Frontend: React / Next.js (to implement PWA and Telegram Mini App). • ML/CV: OpenCV, PyTorch, YOLO. • Infrastructure: Docker, deployment on GPU servers (NVIDIA), MQTT. 5. Expected results (Acceptance criteria) 1. Automation: The box status changes in the database within 5 seconds after a physical event without human intervention. 2. Accuracy: Stable recognition of RC numbers (including dirty numbers) with an accuracy of at least 90% in the daytime/evening. 3. Fault tolerance: The application works correctly in the browser when the messenger is blocked. 4. Synchronization: The standby time in the user interface is updated in real time. 6. Milestones of development 1. Stage 1: A prototype of license plate recognition and auto detection on 2 streams. 2. Stage 2: Logic of the "smart queue" and predictive algorithm (Backend). 3. Stage 3: Development of the PWA interface and integration with the Telegram Mini App. 4. Stage 4: Field testing at the facility and finalizing the notification system. 7. Requirements for the code and architectural background • Git: Transfer of the source code to the customer's repository. • Documentation: Mandatory documentation of the API (Swagger/Redoc). • Scalability: The architecture should allow scaling up to 50+ cameras. • Mobile Readiness: The Backend should be designed as a universal RESTful API, completely separate from the frontend logic, to ensure seamless integration with native mobile applications (Flutter/React Native) at the next stage of project development.

1. Full automation of accounting for the employment of posts (excluding the human factor). 2. Increasing the capacity of the facility through predictive time calculation (ETD) and smart queue. 3. Transparent analytics of attendance and service time for the business owner. 4. Minimize queue conflicts through automatic customer notifications.

Alexander Meshcheryakov