Beginning Anomaly Detection Using

seed: An integer to use as a random seed.

 Flatten

keras.layers.Flatten()

This layer takes all of the inputs and flattens them into a single dimension.

Images can have three channels if they’re color images. They can be RGB (red, green, 

blue), BGR (blue, green, red), HSV (hue, saturation, value), etc., so the dimensions of 

these images are actually (height, width, channels) if it’s formatted channels last or 

(channels, height, width) if it’s formatted channels first. To preserve this formatting, 

there is a parameter you can pass in to the flatten layer:

• 

data_format: A string that’s either ‘channels_first’ or ‘channels_last’. 

This tells the flattening layer how to format the flattened output to 

preserve this formatting.

To get a better idea of how the layer flattens the input, check out the summary in 

Figure 

A-14

 of a convolutional neural network.

Appendix A   intro to KerAs

---

333