In this project, I use a LeNet-5 convolutional neural networks to implement a classifier to distinguish German traffic sign. LeNet-5 is the most basic structure of the CNN.
After the conv layer and the pooling layer are repeated twice, fully connected connection layer is used three times.
In the given file, load the data in the following format
Number of training examples = 34799
Number of validation example 4410
Number of testing examples = 12630
Image data shape = (32, 32, 3)
Number of classes = 43
visualize an image about each label and represent the ratio of the element number of the each class to total.
Convert the image to gray color and normalize it for better learning.
By using tensorflow, conv layer, pooling layer, activation, and fully connected layer are implemented. Then compares the calculated logit with the actual label and optimizes it to minimize the difference.
The model is trained by passing the input sign through a convolution neural network and minimizing the difference between the predicted value and the actual value.
Apply the six German traffic signs received from the Internet to the learned models to ensure that the learned models correctly distinguish traffic signs.
contains a complete code to implement the actual project In addition, the results and descriptions of each process have been added to confirm the entire process
describes the important process in detail
converts ipynb file to html
contains new images obtained from internet
contains what the German traffic sign label means
contains parameter values of the trained model
-
Reference : https://github.com/netaz/carnd_traffic_sign_classifier.git
I referenced this link to visualize the data.