Embedded Files
Digit Recognition
Digit Recognition
An image recognition CNN engineered to recognize handwritten digits in an embedded environment, deployed on a Raspberry Pi.
What?
What?
Convolutional neural network engineered to recognize handwritten digits
Trained and tested using the MNIST dataset on a PC with an NVIDIA RTX GPU
Deployed on a Raspberry Pi with a camera, where image processing was performed before classifying the digit
How?
How?
Create the architecture of the neural network using Keras and TensorFlow, written in Python
Configure various hyper-parameters, such as epoch, batch size, and learning rate, to manage over-fitting and under-fitting of the neural network
Utilize image pre-processing techniques to transform the captured camera image to match the black-background and white digit from the MNSIT dataset
Display the digit classification onto an LED display connected to the Pi
Result
Result
Over 90% accuracy on physically hand-written digits on a sheet of paper
Continuous video streaming from the Pi camera to the neural network model, enabling the user to easily identify multiple numbers
Google Sites
Report abuse