Recently, Minivision Technology officially released a new product - Minivision Intelligent Vision Hub Platform, which provides large-scale and centralized view data analysis capabilities for smart cities, fully activating the city's "visual nerves" and helping the city move towards "one network management".
Nowadays, the third wave of information technology revolution has swept the world. Technologies such as 5G, big data, AI, and blockchain have led us into the era of "Internet of Things", and an ideal blueprint for "Internet of Things" is about to emerge.
Cities are beginning to accelerate the pace of digital transformation. The 14th Five Year Plan proposes to use digitalization to promote urban and rural development and innovate governance models. The 20th National Congress report calls for "building livable, resilient, and smart cities".
1、 What are the difficulties in the digital transformation of cities?
The demand for intelligent urban governance continues to be strong, but data, algorithms, and computing power have become the three major pain points of urban digital transformation.
1、 It is difficult to collect massive data in cities, develop algorithms, and incur high operating and maintenance costs.
2、 The algorithm has poor accuracy, does not meet practical needs, and has poor application results.
3、 GPU/NPU hardware costs are high, computing power configurations are not flexible, construction costs are high, and computing power resources are scarce.
2、 Product Architecture
The Minivision Intelligent Vision Central Platform integrates algorithm production, algorithm management, view data analysis, algorithm data services, and other functions. It has advantages such as strong concurrency ability, accurate analysis and recognition, fast computing speed, and low operating costs.
The platform has core functions such as algorithm annotation training, data warehouse, access adaptation, algorithm management, intelligent scheduling, operation and maintenance management.
3、 Weak supervision of small samples and efficient production of algorithms
Through the "small sample weak supervision technology", the algorithm annotation training platform can fully reduce the dependence of algorithm production on large sample data while ensuring algorithm accuracy. From training, analysis to deployment, the entire process of codeless training reduces the production threshold of the algorithm, improves production efficiency by more than 6 times, and significantly improves the speed of algorithm iteration and accuracy. The model upgrade can be achieved once a week.
This means that the platform can provide diversified algorithm customization services, meeting the long tail needs of urban governance. If the government proposes a demand for customized algorithms, Xiaoshi can implement the algorithm within one month at the earliest.
4、 Rich algorithms, full scene coverage
The platform supports 140 mature algorithms+X customized algorithms. The large capacity algorithm warehouse can cover a variety of scenarios under urban governance, such as urban governance, road traffic, environmental protection, smart communities, smart campuses, bright kitchens, construction sites/factories, mines/mines, etc.
At the same time, the algorithm has fast recognition speed and high accuracy, and can also be applied for and freely allocated according to needs; Open API interface for easy integration with third-party algorithms.
5、 Resource pooling and load balancing
Minivision Intelligent Vision Central Platform builds a shared computing resource pool to achieve unified management and intelligent scheduling of computing power. It can also dynamically adjust based on algorithm properties and resource usage.
By combining scheduling strategies such as video round robin and tidal scheduling, the platform can increase the utilization of computing resources by more than 50%, and improve the comprehensive feedback of algorithm task results by more than 35%. In this way, the platform is more able to access millions of channels of video, thereby undertaking city level video analysis tasks.
6、 Build a closed-loop system to reduce costs and increase efficiency
The platform collects urban monitoring data and issues analysis tasks based on government management needs.
In response to emergency events, timely detection, intelligent alarm, closed-loop handling by various departments, and cross departmental collaboration are achieved. Not only that, breaking the perception barriers between commissions and bureaus, allowing data resources to be shared and shared, can significantly improve government governance efficiency and assist in efficient decision-making in urban governance.
7、 Implementation examples, 10+urban construction experience
In a certain district of Jiangsu, Xiaoshi provides intelligent community AI service capabilities that cover cloud, edge, and end. In a certain city in Liaoning, Minivision is constructing an urban governance project for the Urban Management Bureau to achieve AI intelligent urban inspection. The Minivision Intelligent Vision Central Platform has also been widely implemented in multiple provinces and cities such as Guangdong, Shanxi, Anhui, Zhejiang, Gansu, and Shanghai, quietly integrating into local aspects of urban governance, injecting AI vitality into urban development.
