laitimes

From passive capture to active service, how will AI change the surveillance camera market?

author:Titanium Media APP

According to data from Lotu Technology, the sales volume of China's consumer-grade surveillance camera market will reach 53.43 million units in 2023, a year-on-year increase of 10.9%, of which the sales volume in the online market will be 26.63 million units, accounting for 49.9% of all channels. Xiaomi, Joan, and EZVIZ are among the top three in the Chinese market, and a "clairvoyant" that grows at home is becoming the choice of more young people.

Jiang Xin, who is in Beijing, recently experienced a small family crisis, but fortunately, the smart camera in his home has become his savior. Jiang Xin shared this story with the Titanium Media APP: During his business trip, the water pipe in the bathroom at home suddenly burst, and it was the camera at home that captured this scene in time, so that he was able to remotely monitor and take quick measures to avoid the tragedy of turning his home into a sea.

Jiang Xin recalled: "At that time, I was observing the pet cat at home through the camera, and suddenly I heard an abnormal sound of water. If it weren't for the camera, I probably wouldn't have realized there was an emergency at home. This experience made him deeply appreciate the importance of smart cameras, and he even believes that if the camera can actively identify anomalies in the room, it will become a must-have device for young people who live alone.

With the popularization of the Internet, massive information exchange is happening every minute and every second, in the real-world information data, communication elements are collected through a camera, and audio and video have become an important information carrier in modern society, which is used by more and more industries.

However, as more devices and information increase, so does redundant information and noise. How to filter the existing information and then transform the effective information into valuable efficiency improvement for users will be a compulsory course for each product.

At this point in time 2024, AI is clearly the best solution. Collecting data, learning, feedback, and realizing the continuous improvement of AI have become the driving force for surveillance camera companies such as EZVIZ.

From passive capture to active service, how will AI change the surveillance camera market?

Cheng Zhanzhan, head of the algorithm department of EZVIZ, said that AI should have the ability to perceive like a human, the ability to execute like a human, and the ability to think like a human to realize autonomous interaction with the physical world. In addition to capturing and identifying physical signals through the camera, the device also needs to be able to think, for example, the camera must be able to understand whether there are safety risks in the video, such as the elderly falling, children climbing the window, etc.; Robot products need to be able to plan behaviors that correspond to decisions, such as when a family member has eaten, a sweeper can autonomously clean a restaurant.

Cheng Zhanzhan believes that with the rapid development of AI in the past two years, although AI will not be as self-aware as humans, it is enough to change the existing business application mode with the super data and knowledge compression ability of large models. In layman's terms, "for scenario-based AI applications, the current capabilities are sufficient."

Specific to the subdivision of AI, what EZVIZ currently needs to do is to have embodied AI capabilities.

Specifically, there are multiple levels from L0 to L4. It includes simple judgment, communication, interaction, and comprehension skills. When AI capabilities reach L3, AI can actively interact and understand, and can transcend time, space, and environmental elements. For example, when the weather is bad and the owner wants to go out, the AI housekeeper takes the initiative to remind the owner to bring an umbrella; When someone in the family coughs, the AI doctor assistant reminds health-related matters, and has the ability to act as a role-based AI agent.

At the L4 level, AI requires the device to have the ability to act autonomously. When the owner sits on the sofa, the AI robot can autonomously brew a cup of latte and deliver it to the owner according to the owner's habits. Cheng Zhanzhan believes that L3 is achievable at present, and L4 needs to work together with the industry.

From passive capture to active service, how will AI change the surveillance camera market?

In order to better realize the ability from basic event perception, to conversational device interaction, to embodied interaction, and to realize the all-round interaction ability with the physical world, EZVIZ has independently developed a "blue ocean model" for Internet of Things scenarios.

EZVIZ Blue Ocean Large Model is an embodied intelligent large model for IoT scenarios, with three systematic AI capabilities: development and openness, basic interaction and embodied agent, covering the whole link from device perception to device execution, with a variety of devices that can be accessed, capable of perceiving world spatial information, processing multi-modal data and accurately identifying and understanding, and finally planning and guiding the device to respond.

The device-cloud collaboration algorithm capabilities provided by EZVIZ include 73 basic perception algorithms, covering text, audio, image and video, such as face recognition, humanoid detection, OCR, etc., which can be directly called and used for developers to make corresponding AI solutions. For interactive or generative algorithms, EZVIZ also provides 26 capabilities, covering language interaction, image generation, multimodal interaction, and a variety of embodied proxy services.

From passive capture to active service, how will AI change the surveillance camera market?

Compared with home scenarios, image recognition + AI applications in commercial scenarios may be more immediate and can bring benefits to enterprises. At present, EZVIZ has covered scenarios including smart hotels, smart farming, AI convenience stores, and unmanned car washes. Xu Haibo, vice president/CMO of KO billiards, told Titanium Media APP that with the AI+ vision solution provided by EZVIZ, the billiards hall has a completely different way to play. At present, KO billiards has covered more than 300 cities, 5,000 stores and 8 million users.

Based on the video content production and user live broadcast functions of EZVIZ Cloud, KO Billiards can automatically record the user's full video of the whole game, and support live broadcast push, replay or AI automatic editing of the user's wonderful video of clearing the stage in one shot, which is convenient for users to share by themselves and improve user stickiness. Also based on EZVIZ's visual recognition and AI capabilities, KO Billiards can automatically identify users who clear the table at one stroke, which is convenient for the store manager to give drinks or table fees, which is used to help the store operate for a long time.

Specific to the business model and advantages of existing products, EZVIZ said that compared with other 2B companies that focus on software or hardware profit models, EZVIZ emphasizes flexibility and moderation, which can be flexibly adjusted according to the needs of users. Flexibility at the product level is also the advantage of EZVIZ, because the underlying camera equipment is completely self-developed, compared with other companies that provide equipment from suppliers, more parameters can be adjusted, and the sensitivity and power consumption of the equipment can be adjusted according to actual needs, so as to achieve products that meet more requirements.

In view of the latest trends in the application of millimeter-wave radar in the field of smart home by Xiaomi and Huawei, EZVIZ said that the future of smart home must be multi-dimensional and multi-perceptual, and millimeter-wave radar itself is also a good technology, and related work is currently underway. (This article was first published on Titanium Media APP, author|Wu Honglei, editor|Zhong Yi, Jiang Xin is a pseudonym in the article)

Read on