this project is wrriten by python notebook on google colab.
Based on this paper, we want to reduce the false error rate in the process of traffic sign classification.
-
Detection: to simplify the question, we use nueral network to extract the possible signs. we use faster_rcnn_resnet_101
-
Classification: we will try to improve our result. We will implement some popular classification algorithms and then compare them with our method.
we will use German Traffic Sign Recognition Benchmark GTSRB for training and testing.
We will not use neral network methods in classification stage.