A Start-to-Finish Guide to Building Deep Neural Networks in Keras
Everything from image augmentation to plotting accuracy
Learning deep learning is daunting; so libraries like Keras that make it easy are helpful. In this article, I outline, explain, and provide code for 7 steps in building an image recognition deep convolutional neural network in Keras.
1 | Loading Image Data and Basic Preprocessing
Images will (most of the time) be in a .png
or .jpg
format. They can be loaded using the cv2
library with image = cv2.imread(file_directory)
.
The cv2
library has handy exporting from a cv2
image to a numpy array, done through img = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
. This will yield an array of dimensions (m
, n
, 3), where m
and n
are the dimensions of the image. 3 is representative of the depth, or the amount of red, green, and blue to incorporate for the final pixel color.
Finally, the data should be scaled, or put on a scale from between 0 to 1. This improves model performance (mathematically, neural networks operate better on a 0-to-1 scale). This can be done…