The 7th Digital China Construction Summit • Intelligent Computing Cloud Ecology Conference was recently held in Fuzhou. At this 2024 Zhi Suan Cloud Ecology Conference themed "National Cloud Intelligence, Gathering Strength to the New," China Telecom shared and showcased the practice and achievements of industry "AI+" technological innovation with customers and ecological partners, gathering data to generate intelligence, and supporting high-quality development with a new ecosystem of intelligent computing.

Seizing the new opportunities of Digital China and seizing the commanding heights of future development, on May 24, at the "Zhi Suan Cloud Ecology Conference • Industry AI+ Forum," China Telecom and YITU TECH. jointly released the "Emergency Early Warning Vision Large Model." This large model deeply mines the value of emergency industry data and further enhances the intelligent level of risk monitoring and early warning, truly solving the difficult problems in video monitoring and early warning.

Video monitoring and early warning face many difficulties in practical emergency applications, such as the difficulty of retrieving massive emergency video data from multiple sources; difficulty in obtaining training samples; single identification ability, long customization cycle, and high costs. The emergency early warning vision large model jointly created by China Telecom and YITU TECH. fully leverages the strengths of both parties and demonstrates strong practical application capabilities:

• Fast Algorithm Accuracy Improvement

The emergency early warning vision large model is based on YITU Technology's many years of technical accumulation in the field of computer vision. In the zero-sample case, the accuracy can reach 60%. After rapid training with a small sample and continuous fine-tuning training, the risk identification accuracy can quickly exceed 95%.

• Low Overall Input Cost

The hardware foundation of the emergency early warning vision large model adopts a server architecture with memory as memory, and integrates a computing power solution with low-cost inference cards. The total cost is only 1/4 of similar products in the market.

• Low Model Training Threshold

The emergency early warning vision large model can provide users with an integrated training and pushing capability platform. Customers can train with a small number of samples, fine-tune in minutes, and quickly iterate algorithms. Forming their own algorithm library and continuously iterating accuracy. It greatly reduces the requirements for professional personnel and the amount of samples in model training, truly achieving that everyone can be an algorithm engineer.

At the same time, China Telecom fully leverages its cloud and network advantages, deploying the emergency early warning vision large model in a one-stop hierarchical manner. The open training platform can support customers in "self-training," accumulating a vast algorithm library, and continuously iterating accuracy.

The emergency early warning vision large model fully mines the industry value of emergency video data, endowing each video with real intelligence capabilities, and will bring a strong new impetus to the overall improvement of the intelligent level of the emergency system.

With "Artificial Intelligence +" officially becoming a national policy, a brand new intelligent era is accelerating, also providing new historical opportunities for the construction of smart emergencies. China Telecom and YITU TECH.'s cooperation on the "Emergency Early Warning Vision Large Model" empowers smart emergencies with scientific and technological innovation, hoping to work with all sectors of society to jointly expand the application of visual intelligence in industry fields and promote industrial innovation and development together.

You can copy the link to share to others: https://www.yitutech.com/node/916