haseebiba/traffic_sign_classification

traffic_sign_classification

traffic sign classification is an important task for self driving cars. this project is CNN based , using keras and tensflow as the lower layer with 90 percent accuracy.

prerequisite

familarity with python and artifical neural network and basic undertandings.

dataset

The dataset is consist of 43 different classes the image are 32X32 pixels RGB images the data set is beign broken down like 60% training 20% testing 20% validation

intallation

1. Clone this reposiroty or simply download it and put these files into you anaconda directory if using anaconda other wise to the place of your IDE use its saved files.
2. Following directoires which are very important to run this program:

1)matplotlib.pyplot (pip install matplotlib)

2)numpy (pip/pip3 install numpy)

3)tensorflow (pip install --upgrade tensorflow)

4)pandas (pip install pandas)

5)seaborn (pip install seaborn)

6)pickle (pip install pickle)

7)random (pip install random)

ones these all are done run the code.

Example of images of dataset into 10X10 matric of images

Example after normalize

  # converting to grey scale
  X_train_gray= np.sum(X_train/3,axis=3 ,keepdims= True)
  X_test_gray= np.sum(X_test/3,axis=3 ,keepdims= True)
  X_validation_gray= np.sum(X_validation/3,axis=3 ,keepdims= True)

 #normalizing the images 
 X_train_gray_norm=(X_train_gray -128)/128
 X_test_gray_norm=(X_test_gray -128)/128
 X_validation_gray_norm=(X_validation_gray -128)/128

Training model

This is how you have to make your CNN model with all these parameters according to the images and the number of classes

from tensorflow.keras import datasets, layers, models
CNN= models.Sequential()

CNN.add(layers.Conv2D(6, (5, 5) , activation = 'relu' , input_shape = (32, 32, 1)))
CNN.add(layers.AveragePooling2D())

CNN.add(layers.Dropout(0.2))

CNN.add(layers.Conv2D(16, (5, 5) , activation = 'relu' ))
CNN.add(layers.AveragePooling2D())

CNN.add(layers.Flatten())

CNN.add(layers.Dense(120, activation='relu'))
CNN.add(layers.Dense(84, activation='relu'))
CNN.add(layers.Dense(43, activation='softmax'))
CNN.summary()

Training and validation loss and accuracy image view

Testing model

the model is being tested over the test data which were given in the data set. following is the code and ouput of the model in the shape of 5X5 image matrix with prediction and real value

L = 5
W = 5

fig, axes = plt.subplots(L, W, figsize = (12, 12))
axes = axes.ravel()

for i in np.arange(0, L*W):
   axes[i].imshow(X_test[i])
   axes[i].set_title('Prediction = {}\n True = {}'.format(predicted_classes[i], y_true[i]))
   axes[i].axis('off')

plt.subplots_adjust(wspace = 1)