What are the steps neccesary to detect traffic lights using machine learning?

asked 2017-09-20 02:14:29 -0500

azgra gravatar image


as title says, I would like to use machine learning to detect traffic lights and its state (red, orange, green). I am pretty much newbie in this field, so what do I have to do?

Do I have to do some detection/template matching before I feed neural network with images and theirs ground truth?

Then neural network can only tell me if there is traffic light in this image, but not its location am I right? So if there is traffic light, I would use template matching to find its location?

Does this what I just wrote make any sense or I am far from reality?


edit retag flag offensive close merge delete



there ARE networks, that do both detection and identification, e.g. ssd or yolo

also: https://github.com/opencv/opencv_cont...

berak gravatar imageberak ( 2017-09-20 02:22:19 -0500 )edit

Thanks, I will look on it.

azgra gravatar imageazgra ( 2017-09-20 11:41:53 -0500 )edit