3 ms·
I work for a company in this space. Our product tracks and describes things of interest (faces, text, moving objects), and allows the user to search for simila
by prospero 18y ago
I work for a company in this space. Our product tracks and describes things of interest (faces, text, moving objects), and allows the user to search for similar items at other times and on other cameras. It doesn't make security guards obsolete by any stretch, but it can make reviewing past footage much less tedious.
The whole "intelligent video" industry is in its infancy. An interesting point, though, is as these analytics improve, so does our ability to limit the scope of video that can be watched. Irrelevant footage can be made off-limits, faces or license plates obscured, etc.
It's not going to happen overnight, but the trend is clear: surveillance video is going to become more structured as time goes on. But this can be used to uphold personal rights as easily as it can be used to bypass them. Laws will just have to adapt in the face of technology.
- khafra 18y agoI'm afraid it's much easier to bypass personal rights than to uphold them. Even though the technology for either is there, the panopticon is a design pattern that keeps occurring in political space. I'm curious, though--does your company do the kind of thing covered by Steve Rambam in his talk at The Last HOPE? ie, recognition of activities as well as of faces, text, etc.? http://www.thelasthope.org/media/audio/64kbps/Featured_Speaker_-_Steven_Rambam_ http://www.thelasthope.org/media/audio/64kbps/Featured_Speak...(Part_1).mp3 http://www.thelasthope.org/media/audio/64kbps/Featured_Speaker_-_Steven_Rambam_ http://www.thelasthope.org/media/audio/64kbps/Featured_Speak...(Part_2).mp3
- prospero 18y agoActivity is an awfully vague term. I know of one company that's hired professional actors to act out different scenarios, in the hopes that they can distill the essence of suspicious activity, and thereby automatically detect it. I'm not optimistic about their chances. However, it's relatively straightforward to detect people tailgating at security entrances (using one passcard to let two people in), or walking the wrong way into the exit gate at an airport. In both these situations, though, the environment is controlled, and the difference between proper and improper behavior is well-defined. Neither of these things are true on, say, a street corner. Human interaction and body language is complicated. I expect we'll have a good way of automatically detecting behavior sometime around when we have general artificial intelligences, and not before.