Convolutional Neural Networks (கண் பார்வை / Vision Networks)

Let's scan the big Pongal Kolam! (பொங்கல் கோலத்தை பார்ப்போமா!)

Technical Meaning: சுருள் நரம்பியல் வலையமைப்புகள் (Surul Narambiyal Valaiyamaippugal) - Neural networks specialized for analyzing images.

The Core Idea

A Convolutional Neural Network (CNN) is a specialized AI model designed specifically for Computer Vision (understanding images). Instead of looking at an entire image at once, a CNN slides a small "filter" over the image, pixel by pixel, to detect simple patterns like edges, curves, and corners.

The Origin Story

In 1998, Yann LeCun developed LeNet-5, a pioneering CNN used to read handwritten zip codes on envelopes for the US Postal Service. The breakthrough was realizing that an image is just a grid of pixels, and a sliding filter (convolution) is the most mathematically efficient way to find local features regardless of where they appear in the image.

The Tamil Analogy

Imagine trying to understand a massive, complex Kolam (கோலம்) drawn in front of a house during Pongal.

If you just stare at the whole Kolam from far away, it's overwhelming. Instead, you take a small Magnifying Glass (the Filter/Kernel) and scan it block by block.

• In the first scan (Layer 1), your magnifying glass detects simple things: straight lines and curves.
• In the next scan (Layer 2), you combine those lines to detect bigger shapes like petals and stars.
• By the final scan, your brain pieces all the petals together and recognizes, "Ah! It's a Lotus Kolam!"

A CNN does exactly this. It slides a mathematical magnifying glass over an image, building up understanding from simple edges to complex objects.

Try It Yourself