(vlog (2025 09 22) "LeNet-5 in Action: Watch the Classic CNN at Work!")

Описание: Ever wondered how a classic Convolutional Neural Network (CNN) like LeNet-5 "sees" things? This video showcases a unique interactive project where you can draw digits and watch LeNet-5 try to recognize them in real-time!

What makes this demo special?

LeNet-5: A Milestone in AI History: LeNet-5 was the first successful CNN for handwritten digit recognition, and its architecture laid the foundation for modern deep learning. We've implemented this classic network entirely in LÖVE 2D so you can experience its power firsthand.

The Magic of Convolutions: See how the network uses convolutional, pooling, and fully connected layers to progressively extract features from an image. The demo provides intuitive visual feedback, showing you how each layer transforms raw pixels into abstract features, ultimately leading to an accurate prediction.

Real-time Interaction & Insight: Draw with your mouse and get instant feedback on the network's predictions and confidence levels for each digit. This project doesn't just show the final result; it visualizes the model's inner workings in a way you've never seen before.

Lightweight & Efficient: The entire LeNet-5 model, with all its weights and biases, runs efficiently within the lightweight LÖVE 2D framework. It's a testament to what's possible with game development tools and creative coding.

In this video, you'll see:

A live demonstration of drawing different digits and watching the network recognize them in real-time.

An interesting look at how the network "hesitates" or makes predictions when faced with ambiguous or poorly drawn digits.

How to Play:

Left Mouse Button: Draw on the canvas.

Right Mouse Button: Erase what you've drawn.

Hold Middle Mouse Button: Drag to move the canvas.

Press 'C': Clear the entire canvas.

Press 'Esc': Quit the application.

Get the Demo Here:
[0uuu.itch.io/handwritten-digit-recognizer-lenet5]

You can get the complete source code and model by unpacking the .love file.

Thanks for watching!

Please subscribe for more content.

:3