Classifying images with PyTorch and Keras
In this section, we’ll try to classify the images of the CIFAR-10 dataset with both PyTorch and Keras. It consists of 60,000 32x32 RGB images, divided into 10 classes of objects. To understand these examples, we’ll first focus on two prerequisites that we haven’t covered until now: how images are represented in DL libraries and data augmentation training techniques.
Convolutional layers in deep learning libraries
PyTorch, Keras, and TensorFlow (TF) have out-of-the-gate support for 1D, 2D, and 3D convolutions. The inputs and outputs of the convolution operation are tensors. A 1D convolution with multiple input/output slices would have 3D input and output tensors. Their axes can be in either SCW or SWC order, where we have the following:
- S: The index of the sample in the mini-batch
- C: The index of the depth slice in the volume
- W: The content of the slice
In the same way, a 2D convolution will...
