The University of California- San Diego has developed a pedestrian detection algorithm for driving cars. This new algorithm is quicker and more accurate than the existing systems that we have in place now. In fact, it can spot people on the street as fast as we can; however, there are a few drawbacks.
To begin, in order to achieve this amazing feat, the researchers had to first focus on eliminating the areas that do not contain people, such as the sky and the empty roads. Next, they had to lean on deep learning; specifically, they turned to complex image recognition.
Obviously, the system has to know what it is looking at so that it can take the correct actions. It saves a lot in computing power because the focus is limited to specific areas and not the whole picture.
Right now, the new algorithm can only recognize one object a time. Obviously, that is a bit of a downside. Still, detecting people as fast as we can is impressive; now, we just need it to detect as many things as we can.
To that end, the team who developed the tech has plans to perfect the system in order to allow the detection of multiple objects. They also plan to expand the use of this system to other vehicles, robots, and even security cameras.
In this respect, this new development will definitely help us in terms of safety. We can hope for better, more efficient ways of catching people doing crimes, and also prevent accidents from happening. And obviously, for self driving cars, this will increase the safety features.
Here is a sneak peak at how it works: