Iterations are the number of steps through batches of the training data needed to complete one epoch.
For better understanding, consider the following example:
The total training set of images = 5000
Batch Size = 64
From the above values, 64 samples will be taken each time through the neural network forward and backward. This constitutes one iteration.
Therefore, if you are loading 64 images at a time (in one batch), to go through all 5000 images, you will need 5000 / 64 = about 78 iterations.