The surveillance technology industry is in the spotlight today, but that’s not the only reason. Controversy over U.S. Immigration and Customs Enforcement’s use of Flock’s camera network to surveil people, and home camera maker Ring drawing criticism for developing a new feature that allows law enforcement to ask homeowners for footage of their neighborhoods, are currently sparking a broader debate about safety, privacy, and who can monitor who.
But the controversy won’t kill the market, and continued improvements in vision language models are only giving more wind to companies building new ways to help them monitor what’s going on within their facilities.
Matan Goldner, co-founder and CEO of video surveillance startup Conntour, says his company is very careful about which customers it sells to because ethics around the subject are so important. While this may not seem like sound business sense for a startup that’s only two years old, Goldner says Conntour can afford to do this because it already has several large government and listed customers, one of which is Singapore’s Central Drug Enforcement Agency.
“The fact that we have such a large customer base allows us to continue to choose and control our customers.” […] We actually control who uses it, what use cases there are, and we can choose what we think is moral and of course legal. We make all the decisions and decide whether we’re okay or not based on a particular customer. [to work with] Because we know how they use it,” Goldner told TechCrunch in an exclusive interview.
This traction helps Conntour with more than its selective power. Investors are taking notice: The startup recently raised $7 million in a seed round from General Catalyst, Y Combinator, SV Angel, and Liquid 2 Ventures.
Goldner said the round was completed within 72 hours. “I think we had about 90 meetings scheduled over about eight days, and three days later, we started on Monday and were done by Wednesday afternoon,” he said.
In any case, Conntour may be right to be picky, especially given how powerful AI tools in this field have become. The company’s proprietary video platform uses AI models to enable security personnel to use natural language to query camera feeds and search for any object, person, or situation in footage in real-time. It’s a Google-like search engine made specifically for security video feeds. You can also independently monitor and detect threats based on pre-configured rules and automatically display alerts.
Unlike traditional systems that rely on preset definitions and parameters to detect specific objects, movement patterns, and actions, Conntour claims that its system uses natural and visual language models, resulting in a high degree of flexibility and ease of use. When a user asks, “Find the person in sneakers to hand me a bag in the lobby,” Conntour’s system quickly searches all recorded or live video feeds and returns relevant results.
The platform also has built-in AI models that allow users to simply ask questions about footage and receive text answers along with relevant video feeds, as well as generate incident reports.
However, the company’s selling point is its scalability. Goldner explained that the platform is primarily different from other AI video search services because it is designed to scale efficiently to systems consisting of thousands of camera feeds. In fact, he said Conntour’s system can monitor up to 50 camera feeds from a single consumer GPU like Nvidia’s RTX 4090.
The company accomplishes this by using multiple models and logic systems and identifying which models and systems the algorithm uses for each query to minimize the computing power needed to provide the best results to users.
Conntour claims its systems can be deployed entirely on-premises, entirely in the cloud, or a combination of both. It can connect to most security systems you already have or act as a complete monitoring platform on its own.
However, there are long-standing issues in the video surveillance industry. That is, the quality of surveillance depends on the footage captured. For example, it’s difficult to make out details from dimly lit parking lot footage recorded by a low-resolution camera with a dirty lens.
Goldner says Conntour avoids this necessity by providing a confidence score along with search results. If the camera feed source is not of sufficient quality, the system returns results with a low confidence level.
Going forward, Goldner said the biggest technical problem to be solved is bringing full-level LLM functionality to the system while maintaining efficiency.
“There are two things we want to do at the same time, and they conflict with each other. On the one hand, we want to give you full LLM-style natural language flexibility so you can ask anything, and on the other hand, we want to be efficient, so we want to use very few resources. [thousands] The number of feeds is insane. This contradiction is the biggest technical barrier and technical problem in our field, and we’re working really hard to solve it. ”
Source link
