A group of scientists have developed a new system of AI to observe street signs, which need to be replaced or repaired by Google Street View.
Fully automatic system developed by researchers at a university is being trained RMIT in Australia to use the feature “discovery objects,” supported by AI to identify street signs in the images available for free.
The municipal authorities are currently spending large amounts of time and money in monitoring and recording the geographical location of the traffic infrastructure manually, a task that also exposes workers to unnecessary traffic risks, the researchers said.
According to the study, published in the Journal of Computers, Environment and Urban Systems, the model detects the accuracy of nearly 96 percent, which defined its type at an accuracy of about 98 percent, and can record the precise geographic location of the two-dimensional images.
Andrew Campbell at the University RMIT said that the model is trained to see “stop” and “make room” signs, yet it can be trained to identify many other inputs and is easily expandable for use by local governments and traffic authorities
“The authorities have requirements to monitor this infrastructure but there is no cheap or effective way to do this, but with free and open source tools, we have developed a fully automated system to do that more accurately,” Campbell said.