Is midpoint pooling median pooling, and why no max pooling?

Does midpoint pooling simply imply median pooling? If not, what exactly does it do?

Also, may I request (with low priority) max pooling as opposed to mean pooling. Max pooling is a staple of computer vision algorithms. Just throw it waaaaay back in your to-do list. I would never intend such requests to create any pressure on you.


Midpoint pooling actually does max pooling. Please note that versus creating a lot of algorithms for common use I made standard algorithms have more features through their arguments. See the bias argument of midpoint pooling. You can change it to either do max/min pooling by adjusting the bias argument.

Mean pooling is just the average. Midping is the max - min over 2. When you change the bias it becomes either all max or all min.

That’s more than sufficient. No need to add a function if parameters to an existing function serve the purpose. I just overlooked the feature. Thanks.

Like if you want adaptive thresholding the mean/median/mode/midpoint kernel filters all support this along with a few other methods.

Because memory is limited and I have to avoid copying you’ll find I combined a lot of algorithms into one.