Page 224 - AI for Good-Innovate for Impact Final Report 2024

AI for Good-Innovate for Impact

"application and algorithm" dual decoupling service combination of distributed cluster, based
on cloud edge resource collaboration technology. The video decoding and AI inference service
computing power sinks to the city node, reduces the network pressure of the central node
and the concurrency and delay of the single node of the AI inference platform, realizes the
unified scheduling of video analysis computing power, AI analysis computing power and video
resources at the provincial side, the city edge side and the user side, and achieves the purpose
of efficient allocation of computing resources. Integrating machine vision capabilities such
as video and image OCR, it is oriented to multiple industries, sedimenting industry video
AI analysis capability set, and providing atomized AI capability with rapid cross-scene reuse
capability for industry video AI enablement scenes. Moreover, relying on multi-modal large
model technology and related capabilities of Tuosen, it can identify scenes with high false
positive rate of traditional AI capabilities, such as: Fire identification, illegal discharge, etc., the
second audit of large models has greatly improved the identification accuracy of such scenes.

The project opened a total of 50,000 + algorithms for intelligent video analysis scenarios, and
based on the "video +AI" fusion mode to create universal, high-precision, and rapid scene-
oriented AI video standardized products, to achieve the combined reuse of AI capabilities and
the standardization of video +AI one-stop overall solution. Vertical industries such as key forest
fire prevention, water resources protection, and farmland supervision sites have been applied
in scenarios, creating economic benefits of more than 30 million yuan.

52�2�2 Future Work

In the future, this project will give full play to the natural advantages of China Mobile's "operator
connection value", build the MaaS capability of "platform + ecology", realize the end-to-end
intelligent video product innovation of "platform +AI+ application + business operation", create
a new type of infrastructure capability, integrate different industry scenarios, and empower
thousands of industries. Become an important force to promote industrial upgrading and
improve the quality of life. For example, integrate into urban management and public safety
and other fields, build smart transportation, smart city and other business scenarios, and provide
more comfortable living conditions for urban residents; Integrate industrial manufacturing
and food production, optimize the production process, efficiently supervise the production
environment and production quality and other factors, accelerate the industrialization process
and improve food safety.

52�3� Use case Requirements

• REQ-01: To obtain the video surveillance data from public areas, China Mobile must
obtain the authorization of both the government and administrators.
• NOTE: China Mobile can sign an informed consent form with the government and
administrators, clearly stating the purpose, scope, and confidentiality measures for the
collection of video data.
• REQ-02: To obtain the small model and larger model, China Mobile must obtain the open
source AI algorithms and conduct secondary training and development.
• NOTE: After finding the model on an open source platform or GitHub, download the
model file and code directly. Modify the code of the model to fit your data and tasks as
required by the project. When using open source models, ensure that the copyright and
license agreements for the models are adhered to.

208