"Large models open a new era of intelligent security," on January 19, 2024, the 10th Smart City Construction Innovation and Development Conference was held in the Great China Shangri-La Hotel, Futian District, Shenzhen. Leaders and experts from the field of smart city construction across the country gathered to discuss the technical achievements and innovative applications of large models in the field of intelligent security, and to discuss and exchange comprehensively on how to make smart city construction more efficient and scientific in the era of large models.
YITU has been focusing on using advanced artificial intelligence technology to empower smart city construction for many years, combining image perception, video perception with large models, and taking the lead in practice in various fields of smart cities, promoting the rapid transformation of core technologies into real combat effectiveness and productivity. At this conference, YITU TECH. delivered a keynote speech on "The New Era of Large Models, YITU "QuestMind" LVLM Empowers Intelligent Transformation," and shared the practical experience and achievements of the "QuestMind" LVLM in empowering smart city construction and intelligent transformation.
The "QuestMind" LVLM of YITU TECH., which can converse, think, and evolve, supports massive video semantic searches, supports text-based video searches, and has the ability to search at the level of hundreds of millions of seconds on a single machine. It can quickly and accurately lock the target based on features; relying on YITU's full-stack artificial intelligence capabilities, it can achieve one-stop model training, development, and release, zero-sample cold start, and on-site algorithm iteration, continuously upgrading in actual combat, getting better and better, greatly shortening the model modeling and development cycle, and maximizing video value with the smallest model cost; at the same time, commonly used algorithms for key area scenarios can be built-in to achieve out-of-the-box use; YITU Xiaoming AI Agent can make machines accurately understand user needs, and automatically generate the required algorithms based on simple instructions, truly making everyone an algorithm engineer.
Currently, the smart city construction is transforming from data intelligence represented by big data to knowledge intelligence based on large models. YITU's "QuestMind" LVLM has been deployed in dozens of projects nationwide, fully aiding the construction of new-type smart cities and making smart cities smarter. The live demonstration of the "QuestMind" LVLM's combat capability at the scene has also become the key to its rapid practical application and commercialization:
- Video semantic search capability, supporting natural language search for various videos: natural language expressions such as "cars blocking fire passages," "waterlogged roads," and "unattended red suitcases" can quickly find the corresponding videos, giving videos the ability to semantically search.
- Zero samples, cold start: Create an algorithm for not wearing a helmet while riding a bike, just input "not wearing a helmet while riding a bike," and the algorithm can be immediately generated, and once the algorithm is online, it can be trained online in minutes, and the accuracy will increase with use.
- Built-in a large number of algorithms, condensing YITU's years of industry cognition and practice: by putting specific data into the training model, such as typical algorithms for large passenger flow and key area control, to improve model performance. For example, in urban road sections prone to traffic accidents, YITU's "QuestMind" LVLM empowers the old front-end cameras to solve traffic problems, and can quickly deploy and launch the section's common traffic problems such as: running red lights, speeding, not yielding to pedestrians, not wearing safety helmets while riding, illegal U-turns, and driving over solid lines, etc. The system can quickly go online, and the accuracy can reach nearly 100% one week after deployment, with construction costs significantly lower than traditional solutions.
At the same time, at the general meeting held by the Shenzhen Intelligent Security Industry Association, YITU was awarded the "Top 10 Standardization Enterprises" and "Innovative Product Service Provider" awards for its continuous product innovation and business service capabilities. This is an affirmation of YITU's contributions and achievements in the field of security in the past year, as well as encouragement and expectations for YITU's future development.
"Concept first, benchmarking the world's first-class; practicality is king, cost determines scale!" In the era of large models, YITU is bringing its full-stack technical strength that benchmarks the world's first-class, adhering to the operational concept of "maximizing value, minimizing cost," and with an open attitude, looking forward to working with many industry partners to jointly write a new chapter in the construction of smart cities.
You can copy the link to share to others: http://www.hengtetube.com/node/910
Related articles
-
YITU TECH. Wins the 1st Place in Identification Accuracy In Face Recognition Prize
-
YITU Technology’s Research Scientist Wu Shuang Delivers EmTech Asia Presentation on the Evolution of an AI Startup
-
China’s Yitu Technology wants to ‘raise AI capabilities’ in Singapore, Southeast Asia
-
Based on the YITU QuestMind?LVLM, the Dangerous Traffic Behavior Identification System Officially Launched in Karamay