This AI can search people through height, gender by observation camera

AI has taken a new step when it is possible to identify and search for a person not only through his face but also through his height, gender and clothing colors.

AI researchers in India have successfully developed a special tool to find one or more people in a video from security surveillance cameras based on height, color, and gender.

According to Thenextweb, scientists used deep learning technology to train the convolutional neural network (CNN) to understand how to identify human characteristics (soft biometrics) using Computer vision.

You just need to order and tell the AI ​​about the details related to the person you are looking for. From there, will find by the person who matches your requirements. For example, you ask to look for a woman wearing a red shirt, 1m53 tall. The system will narrow the scope of the search in the video and only search for those who recognize the red shirt.

According to the researchers, the algorithm has been found to be accurate to 28 out of 41 people with complex physical characteristics. Currently, this tool is only searchable by height, clothing color and gender.

This is a special tool for finding one or more people in a video from security surveillance cameras.

At first glance, this tool seems not new, but the application potential of this AI system is huge. The most interesting thing is that the accuracy of AI needs to be further improved, even beyond the ability to judge and identify people.

There are situations where people are almost hard to know what to look for in a real-time video surveillance database. But the test on the CNN tool shows that it is very suitable to use to find the timeline related to a specific object, based on footage from surveillance cameras.

Imagine a situation where a person is missing for more than two days. Authorities will search for footage from surveillance cameras at locations where the person may have passed or came here, such as gas stations, campuses, etc. But in the end, they still could not find Anyone who has the same identity as that person, plus can't determine which scene to watch next because of the large number of video surveillance in 2 days.

However, if we can bring those surveillance videos to a neural network, then narrow the monitoring period, the system can track exactly who you are looking for.

Like any other people-based surveillance and search system, this Indian researchers' tool will surely face many people's objections for personal privacy reasons. But considering the benefits this tool offers, they will "benefit more than harm" , because it can help detect and arrest suspects, wanted criminals or search for clues related to a case. kidnapping, terrorism.

The team is expected to continue to improve this tool more accurately and efficiently to find the right people even in crowded and complex environments.