michaelhtleung / handwritten-number-classifier

Machine learning classifier that recognizes handwritten digits photographed by the Raspberry Pi.

Raspberry Pi Handwritten Digit Classifier

What it does

A machine learning number classifier that recognizes handwritten digits powered by the Raspberry Pi.

The Raspberry Pi photographs images of handwritten digits and sends them to a web server that trains a neural network to classify each image. The neural network's prediction is then displayed on the 7 segment display.

Video

1. Raspberry Pi number classifier demonstration

2. Larson scanner demonstration

Pictures

1. Handwritten digit classifier input image

2. Handwritten digit classifier loading shown with Larson scanner

3. Handwritten digit classifier output answer

4. Hardware close up

Python3 Libraries, Frameworks, and Modules

Server-side

• OpenCV
• TensorFlow
• Flask
• numpy
• os

Client-side

• PiCamera
• threading
• requests
• RPi.GPIO
• os

Parts List

• Raspberry Pi 3 B+
• Raspberry Pi Camera V2.1
• 7 segment display
• 74HC595 shift register
• potentiometer

Concepts Applied

• neural networks
• computer vision
• image processing
• multi-threaded programming (to run Larson Scanner)
• webserver communication
• bit shifting