The future of video security: Poll finds surge in large-scale AI deployment

Increasingly, AI is used in video security for various security and non-security objectives. Security-wise, AI can effectively detect intrusion and other suspicious events. Meanwhile, AI plays a key role in intelligent transportation, preventive maintenance and other non-security applications.
 
Today, various types of AI are used in video security. To boost the power of AI even further, large-scale AI models have emerged. Trained on large datasets, large-scale AI models promise to enhance detection accuracy, lower false alarm rates, and enable natural language-based smart search.
 
To get a further understanding on this rising AI technology, asmag.com and Hikvision have teamed up to conduct a survey examining large-scale AI models in video security. This article focuses on user awareness, adoption rate and the level of importance users attach to large-scale AI models, while a partner piece focuses on pain points addressed by large-scale AI, but also challenges inhibiting its wider adoption.

Adoption rate of large-scale AI models on the rise

 

75% are already adopting or planning to adopt large-scale AI-enabled devices

 
When asked “Are you currently using any products or solutions powered by large-scale AI models for video security,” over half, 55.4%, said yes, while 23.4% said no. Interestingly, nearly 20% said they were not using large-scale AI models at the moment but were are planning to. The responses highlight strong adoption rate and growth potential for large-scale AI models. Indeed, with the growing availability of large-scale AI products, such as Hikvision’s DeepinView X cameras and AcuSeek NVRs, adoption of large-scale AI models is all but set to increase in the near term.
 

Significant user awareness on large-scale AI models

 

Large-scale AI models and their functionalities are often discussed in projects, specs, or client meetings

 
To the question “Have you recently seen large-scale AI models and their functionalities being discussed in projects, specs, or client meetings,” the responses strongly suggest the topic does come up on such occasions. The numbers are especially prominent for “intelligent object detection and classification,” where 57.7% of respondents say large-scale AI models are either “frequently discussed” or “actively influencing decisions.” This is followed by “automated security event detection and reporting” (57.1%) and “improved perimeter protection” (55.4%). The numbers highlight strong user awareness towards large-scale AI models, and discussions over the subject have even influenced users to acquire related products and solutions .
 

Over half confident they understand the difference between large-scale AI and traditional AI

 
When asked if they are confident about understanding the difference between large-scale AI models and conventional AI, 32.5% of respondents said they are “confident” and 15.7% “very confident” about understanding the technological differences between the two. The numbers are 28.3% and 22.9%, respectively, when it comes to understanding the differences between the two in terms of benefits large-scale AI brings to the workflow of security personnel. The numbers suggest a certain level of knowledge about large-scale AI models among users, some of whom are already quite well-versed in the new AI technology.
  

Strong recognition of the importance of large-scale AI models

 

52% consider adoption “very important” for their future competitiveness

 
To the question “How important do you think is adopting the latest large-scale AI models to the future of video security,” a significant 52.4% say it is “very important,” while 25.9% think it’s important. 18%, meanwhile, said it’s “moderately important.” This indicates users attach a greater importance to large-scale AI models, which are no longer viewed as a mere “nice-to-have” but may one day become a core component in video security.
  

Majority see clear operational value, including better efficiency and fewer errors

 
When asked “What value do you see in the emerging deployments or applications of large-scale AI models in video security so far,” a total of 73.4% cited “improved operational efficiency.” This is followed by: reduced human error (62%), better end-user experience (55.4%), and cost reduction (52.4%).
 
Indeed, large-scale AI models can effectively detect abnormalities and issue alerts accordingly, improving efficiency for operators who can focus on more important matters at hand. Large-scale AI models maintain constant vigilance 24/7 without declining performance, avoiding errors that humans may make. Finally, large-scale AI models allow users to conduct search using natural language. For example, with systems like Hikvision’s AcuSeek NVRs, users can input queries such as “Show me all instances of a red car entering Zone A yesterday,” and the system retrieves the relevant footage quickly by matching object features with textual descriptions. This significantly boosts the user experience.
 

Hikvision’s large-scale AI products

 
Hikvision has announced a series of new products powered by large-scale AI models, called “Guanlan.” Below we look at how these solutions are viewed by respondents.
 
When asked “Which large-scale AI-powered Hikvision products are you using or evaluating at the moment,” most users, at 75.6%, cite AcuSeek NVRs. This is followed by 70.6% for DeepinViewX cameras and 68% for traffic cameras/servers powered by large-scale AI. DeepinViewX cameras, powered by Guanlan, can distinguish real threats from harmless activities, such as moving tree shadows, small animals, or bright lights, reducing false alarms by more than 90% and allowing security teams to focus on actual problems. The AcuSeek NVRs, meanwhile, leverage Guanlan multi-modal large-scale AI to establish cross-modal text-image comprehension capabilities – users can simply input a single phrase or keyword, and the system will instantly retrieve related video.
 
For those using or evaluating Hikvision products powered by large-scale AI models, a large majority, 72.9%, gave overall positive feedback. This includes 37.1% who said the products were “very useful” and 35.8% who said they were useful – numbers that give testament to the quality and performance of Hikvision’s large-scale AI-enabled products.
 
As for which Hikvision Guanlan product features respondents are most interested in, over half, 53.5%, chose smart object detection and classification. This is followed by more accurate perimeter protection with fewer false alarms (38.5%), text-based video search (33.5%), intelligent traffic management with much lower false detection rate (32.8%), intelligent alarm setup (30.7%) and large-scale AI deployment across multiple devices (22.8%). The numbers indicate that accurate detection and natural language-based smart search are still highly prioritized by customers.
  

Conclusion

 
In conclusion, the findings clearly demonstrate that large-scale AI models are gaining momentum in video security. With over half of respondents already using large-scale AI and nearly 20% planning to adopt them, the trend is undisputable: large-scale AI is moving from a mere concept to an operational reality. Increased user awareness and confidence in understanding the technology give proof to this phenomenon. Meanwhile, strong recognition towards the value of large-scale AI models, especially in the areas of operational efficiency improvement, human error reduction and user experience enhancement, bodes well for further demand and growth. As the technology continues to evolve, large-scale AI models’ role in video security will only deepen, and end user organizations that embrace this technology will be better positioned to leverage their full potential.
 
To explore the full survey findings and detailed industry insights on leveraging large-scale AI for video security, click here.