Output for tensorflow custom object detection model

Question

Im new to object detection using the tensorflow library. Im following a guide where the model outputs the coordinates for one box, like shown in the code below:

...
# Last three layers
locator_branch = layers.Dense(64, activation='relu')(locator_branch)
locator_branch = layers.Dense(32, activation='relu')(locator_branch)
locator_branch = layers.Dense(4, activation='sigmoid')(locator_branch) # Output

The output on this will be output xmin, xmax, ymin, xmax of the box. How can I instead make it output different amount of boxes? Lets say between 1 and 3 depending on how many objects it can find in an image? What does the last layer have to be then?