What are the numerous filters that are used in cnns

Hello all, I have recently completed the week 1 module in course 4, where Andrew explains the Lenet architecture, which has multiple filters like 6. So I would like to know what are the filters that are used in it, is it horizontal, vertical or is it 45 degrees,
30 degrees etc.

In the same way in some other examples we are using,
(128) filters , so what are those 128 filters !

Is there any way by which we can know what are those 128 filters ? If yes please let me know the resources to it.

Hey, the filters in a CNN are trainable parameters, so those 128 filters depend on the training. They could be anything depending on the training input. Its always fun and insightful to visualise them after training :slight_smile:

Thanks for the response @XpRienzo
Another small question here, why did the author’s decided to take only 128 filters, why not filters with odd numbers. Is there any rule to not to take the filters with odd values.

I’m personally not sure if there’s any rule to take specific numbers of filters.

If you don’t mind, can you list me any more resources regarding this topic if you have any ! Or Else that’s fine. Anyways thanks for your swift response. :slightly_smiling_face: