Monitoring the generator and the discriminator loss

The loss can be monitored in the TensorBoard dashboard. The TensorBoard dashboard can be invoked as follows:

1. From the terminal, run the following command:

      tensorboard --logdir=./logs

./logs is the destination where the Tensorboard logs specific to the program are stored and should be defined in the program as follows:

      self.writer = tf.summary.FileWriter("./logs", self.sess.graph)

2. Once the command in step 1 is executed, navigate to the localhost:6006 site for TensorBoard:

Illustrated in the following screenshot are a few traces of the generator and discriminator losses as viewed in TensorBoard during the training of the DiscoGAN implemented in the project:

Figure 4.2: Tensorboard Scalars section containing the traces for different losses

The following screenshot shows the loss components of the discriminator...