CIFAR-10 Dataset

Overview

  • Training Set: 50,000 training images & labels
  • Test Set: 10,000 test images & labels
  • Color images

Data Format

  • Each 32 x 32 pixel single digit image is represented as a 32 x 32 array, with color (3 channel) values. ranging from 0 - 255.
  • With 60,000 images, each with 3 color channels sized 32 x 32 results in a 4D array:
    • (60,000, 32, 32, 3)
  • Classes
    • 10 different objects
    • airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck
  • Labels
    • List of numbers ranging from 0 - 9.

Documentation

TensorFlow

TF MNIST Dataset

Distributed Training with CIFAR-10 Dataset