LeNet for CIFAR10 Data

Now that we have learned to build and train the CNN model using MNIST data set with TensorFlow and Keras, let us repeat the exercise with CIFAR10 dataset. 

The CIFAR-10 dataset consists of 60,000 RGB color images of the shape 32x32 pixels. The images are equally divided into 10 different categories or classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, and truck. CIFAR-10 and CIFAR-100 are subsets of a large image dataset comprising of 80 million images. The CIFAR data sets were collected and labelled by Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton. The numbers 10 and 100 represent the number of classes of images.  

 

Note

More details about the CIFAR dataset are available at the following links: http://www.cs.toronto.edu/~kriz/cifar.html and http://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf.

We picked CIFAR 10, since it has 3 channels, i.e. the depth of the images is 3, while the MNIST data set had only one channel. For the sake of brevity...