Confused about when to use a Flatten() layer in keras/tensorflow

Hello I am currently learning deep learning in Python using Keras/Tensorflow and am starting my own image classification project. I have looked through a ton of example code online and am very confused about the Flatten() function....

I have seen image classification examples that have a Flatten() layer and others that dont. But the documentation I have read seems to indicated the a Flatten() layer is always neccessary to reshape the image data for output.

Can someone please explain the …

deeplearning flatten keras tensorflow