This page complements the Deep CNN (Convolutional Neural Networks) example on the Convolutional Neural Networks page. This example tutorial is intended for image classification problem. It shows how to use TensorFlow in order to train and evaluate the CIFAR-10 dataset with a CNN. This CNN can be run on both CPU and GPU.
Image classification techniques, in general, requires quite a bit of computational power and training neural networks is computationally expensive. Consequently, training neural networks for image classification requires massive amount of computation.
To tackle this situation, the example script is written to be run not only on CPU, but also on GPU. The script support a single GPU card and multiple GPUS cards as well. (Unfortunately, I don’t have a GPU card. So I ran this script in a CPU-only environment.)