Beginning Anomaly Detection Using

Figure 4-48.  Code to compile the model

Figure 4-49.  Code to start training the model

Chapter 4   autoenCoders

---

158

Now that the training process is complete, let’s evaluate the model for loss and 

accuracy. Figure 

4-50

 shows that the accuracy is 0.81, which is pretty good. It also shows 

the code to evaluate the model.

The next step is to use the model to generate the output images for the testing subset. 

This will show you how well the reconstruction phase is going on. Figure 

4-51

 shows the 

code to display denoised images.

Figure 4-50.  Code to evaluate the model

Figure 4-51.  Code to display denoised images

Chapter 4   autoenCoders

---

159

You can also see how the encoder phase is working by displaying the test subset 

images in this phase. Figure 

4-52

 show the code to display encoded images.

Figure 4-52.  Code to display encoded images

Figure 

4-53

 shows the graph of the model as visualized by TensorBoard.

Chapter 4   autoenCoders

---

160

Figure 

4-54

 shows the plotting of the accuracy during the training process through 

the epochs of training.

Figure 4-53.  Model graph shown in TensorBoard

Chapter 4   autoenCoders

---

161

Figure 

4-55

 shows the plotting of the loss during the training process through the 

epochs of training.

Figure 4-54.  Plotting of accuracy shown in TensorBoard

Figure 4-55.  Plotting of loss shown in TensorBoard

Chapter 4   autoenCoders

---

162

Figure 

4-56

 shows the plotting of the accuracy of validation during the training 

process through the epochs of training.

Figure 4-56.  Plotting of validation accuracy shown in TensorBoard

Figure 

4-57

 shows the plotting of the loss of validation during the training process 

through the epochs of training.

Figure 4-57.  Plotting of validation loss shown in TensorBoard

Chapter 4   autoenCoders

---

163