Manoj Rao bio photo

Manoj Rao

Your Average Common Man

Email Twitter Github


ConvNets might be some boring concepts that are not worth paying too much attention. But, for now, let’s assume they have some sort of advantage over Multi-Layer Perceptrons that make them worth understanding.

Where’s Waldo?

Before you go finding Waldo, just think consciously about how your mind and eyes are working to find Waldo in the image. In this image, Waldo is not the “odd man out”, therefore, moving your eyes somewhere outside the image and squinting won’t help here. You can try this if the object or the person you are spotting has a stark contrast from every other object. But it’s not the case in this image, there are too many stripes.

Where's Waldo

If you found him, great! There are a couple of things that remained true to the way you found him. You scanned the image in smaller chunks and looked for a “pattern”. When you recognize an object you respond to it the same way, irrespective of where you find this object in the image. Of course, scientists put a fancy label on this Translation Invariance. We mentally divide up the image into pieces/chunks. Our eyes focus on the smaller chunks in this image. chunk your eyes/you must rrespective on what’s showing in the neighboring (or a distant chunk), and the fancy terms (or not so much in this case), Locality.

This is the heart of the idea behind ConvNets. We will look at the Maths behind it in the future post.

I was referring [Conv
Nets](, if you have the ability to keep the focus, this book is fantastic. I try to
simplify the material from this source, albeit poorly, here.

My Podcast!

If you like topics such as this then please consider subscribing to my podcast. I talk to some of the stalwarts in tech and ask them what their favorite productivity hacks are:

Available on iTunes Podcast

Visit Void Star Podcast’s page on iTunes Podcast Portal. Please Click ‘Subscribe’, leave a comment.

Get it iTunes