AI of the Moscow region
Problem: Long processing times for requests: each message must undergo manual moderation to determine the subject and performer; Inefficient use of resources: labor-intensive and manual reconciliation of reports with photo verification for violations; Lack of proactive control: if there are information systems, monitoring is carried out manually. Solution: The project is a comprehensive solution based on artificial intelligence (AI) using various AI methods (computer vision (CV), natural language processing (NLP), machine learning (base ML), intelligent decision support) aimed at improving the efficiency of information systems and automating the processing of citizens' requests. Solution architecture: a self-written neural network based on the Python programming language using the PyTorch library is built on the basis of an ensemble of multiclass models for detecting objects in images with high accuracy using tracking logic, text message classification models, a combination of various neural network architectures using an instance segmentation model. Automation of detection and control of violations: AI models analyze images, text and video, allowing you to automatically identify violations and monitor their elimination: analysis of photos of performers in reports on violations: checking the uniqueness and quality of photos, searching for the necessary classes of violations in the photo, marking violations; monitoring of video surveillance cameras: analysis by the AI model of the cropped video stream, finding and marking patterns of violations; classification of the subject of messages from citizens: determining the category of the problem and the specific performer based on the approved classifier. Integration with existing systems: for seamless operation and access to the necessary data, the solution is integrated with eight regional state information systems of the Moscow region - the portals "Dobrodel" and "Incident Management", the video surveillance system "Safe Region", the systems "Control and planning system in the field of road infrastructure", "Mobile dispatch platform", "Management of urban planning activities", "Unified platform for registration in club formations of cultural centers" and the information and analytical system "Unified Regional Management Center". Monitoring and validation: monthly monitoring and validation of the solution according to functional and statistical metrics allows you to guarantee its accuracy, reliability and compliance with the tasks set.
Artificial Intelligence And Digital Services
Artificial intelligence
12 months
Russian Federation, Moscow region
The ability to refine and adapt: The open code of the neural network makes it easy to refine and adapt the solution to specific tasks; The system can be supplemented with new functions and modules to expand its capabilities. Scalability: The solution can be applied in other organizations and regions, provided that task management systems and/or video surveillance systems are available. The modular architecture makes it easy to integrate the system with existing IT infrastructures. Reducing dependence on third-party vendors: Using proprietary software does not require the purchase of expensive licenses. This ensures financial independence and freedom in the development of the system. Uniqueness: An integrated approach combining various AI methods ensures high efficiency and accuracy of processing; The solution has no analogues in the field of public administration. Increasing data privacy: Control over the neural network code guarantees the confidentiality of the processed data.
Absent
Absent
