Traffic Sign Recognition


This is the first finished part of our Self-Driving vehicle project. This part was all about writing a traffic sign detector, which would run on our quite old Laptop with 1-2 FPS. There have been a lot of problems which needed to be solved. But finally the software works and produces usable results. Of course this piece of software is far away from being perfect. However it is still a good baseline to work with and to learn from. If you are interested in how everything started and what concrete problems occured, have a look at here.

You are free to use the whole project for whatever you want it to use. The software is licensed under zLib license.

Source Code (C/C++ Code and Proto-File): StreetNet Code

Trained CNN: StreetNet Caffemodel

Used dataset: Dataset

Some results (Red rectangles show found traffic signs and blue rectangles mean that in this region may be a traffic sign; For more results have a look at Day #9 & #10 of Online Diary):