Hi Eyal,

Thanks for your comment and reading.

You are right in pointing out the error in the conv layer, but it looks like it's not in the description but in the TensorFlow implementation.

LeNet-5 architecture does have 3 convolutional layers as described, but I've not implemented the third conv layers in the code within the article.

I'll update it shortly.

To be more clear the C5 layer in the architecture is also a convolutional layer

Thanks for bringing this to my attention