BriefCam

Twenty-four hours a day, and no one to watch them

A security camera produces twenty-four hours of footage a day, and almost all of it is completely empty. When an incident happens, someone has to sit down and watch. Not just the ten relevant minutes, since no one knows in advance where they are, but everything that came before and after them.

The bottleneck here is human, not technical, and that is what makes it hard. You can add cameras, you can add storage, and you can improve image quality. Each of the three only increases the amount of footage someone has to go through.

The intuitive solution, playing the video faster, simply does not work. At eight times normal speed, a person who crossed the frame in two seconds becomes a quarter of a second, and the eye misses them. Higher speed does not save attention. It wastes it on empty footage and raises the chance of missing exactly the second that made someone sit down to watch. Many investigations break down at this point, and it happens quietly: no one knows what was not found.

Video synopsis: the real idea

The solution developed here is counterintuitive, and that is exactly why it worked. Instead of speeding up the timeline, the system takes it apart.

First it learns the fixed background of the scene: the wall, the sidewalk, the parked car that does not move, the tree that only sways in the wind. Then it cuts each moving object out of the background separately and keeps it as an independent item, with its real timestamp. Then comes the step that looks strange the first time: dozens of objects that appeared at completely different hours are shown side by side, at the same time, in the same frame.

The result is called a "video synopsis," and it makes it possible to review hours of footage in minutes. The viewer's attention goes to the objects, not to empty time, and clicking on any object takes you back to the original moment it appeared. The idea was published as academic research before it became a product, and its logic can be summed up in one sentence: empty time, not pixel quality, was the resource that had been wasted all those years.

From a Jerusalem lab to the market, via Yissum

The technology came out of research by Prof. Shmuel Peleg at the Hebrew University of Jerusalem. The patent was registered through Yissum, the university's technology transfer company, and from there it passed to a company set up to develop the product: BriefCam, founded in 2007 by Peleg, Gideon Ben-Zvi and Dr. Yaron Caspi. The company's headquarters were in Modi'in-Maccabim-Re'ut.

This path, university research that becomes a commercial product through a technology transfer company, is often described in presentations and rarely completed. Here it was completed all the way, including a sale to an international corporation.

The funding it took was much smaller than what is common today. According to IVC data, the company raised a total of $15.6 million over all its years, and its investors included the investment arm of Motorola Solutions, the Aviv fund and OurCrowd. It is a useful reminder that the size of a funding round does not necessarily predict the outcome, for better or for worse.

What else the system can do

On top of the synopsis, a layer was built that turned the recording into something you can query. You can filter by object type, size, color, direction of movement, speed or time spent in a certain area. An investigator can ask for "every white car that drove in this direction between two and four" and get the results right away.

In BriefCam's own acquisition announcement, the capability was described as locating, tracking, extracting and identifying people and objects, their attributes and their behavior, from video. The same layer was used to build real-time alerts and counting: how many people passed, how many cars entered, how long they stayed.

The real change here concerns the role of the security camera. It goes from a recording device that is watched after the fact to a database you can ask questions. Once you can ask and get an immediate answer, the recording stops being an archive and becomes a search tool. By 2018, according to Canon, the technology was in use in more than 40 countries and regions around the world.

2018: Canon buys, and the price that was not published

On 9 May 2018, Japan's Canon announced that it had signed an agreement to acquire BriefCam's shares, subject to antitrust approval. The acquisition was completed on 2 July 2018 and announced the next day.

Precision is needed here, and it is usually left out: Canon did not publish the deal price. The figure usually quoted, about $90 million, is a Globes report citing market sources, not an official figure from the buyer. The difference between the two is not just semantic. It is the difference between a fact you can check in an official document and an estimate by a third party.

The acquisition was part of a broader Canon move into network video solutions, alongside companies it had bought earlier, including Axis Communications and Milestone Systems. Until 2024, BriefCam continued to operate as an independent company within the group, while staying compatible with other vendors' products. That business decision is far from obvious when a corporation buys a technology that exactly complements the products it sells itself.

In 2024 BriefCam was folded into Milestone Systems, a sister company in the Canon Group. Since then it has been a Milestone product, and its website address now redirects to Milestone's site.

A dual-use tool: the efficiency is also the question

The same tool that shortens an investigation from forty hours to forty minutes also makes mass surveillance cheap and practical. This is not a side effect. It is exactly the same property, measured from two sides.

Its deployment by law enforcement, city governments and public facilities in many countries has drawn public criticism over the scope of surveillance. It is a debate that has not been settled and is not likely to be settled soon. Notice its unusual shape: opponents do not claim the technology does not work or that it is wrong. They claim it works too well.

This is the purest expression of what is called dual-use technology. A tool that made watching footage 15% more efficient would not raise any public question at all, because it would not change what can actually be done. It is the engineering success itself that creates the question. This pattern repeats in almost every field where artificial intelligence makes a task that used to require expensive human attention cheaper by an order of magnitude.