MNIST digit classifier which classifies 28 by 28 pixels images into classes: 0 to 9.
The dataset is MNIST and train data has about 60k images as well as labels.
Test data has 10k images.
I trained this data with Sklearn Random Forest Classifier.
Install conda on your machine:
install conda
Verify the installation of jupyter notebook.
clone the repository:
cd to_your_project_path
git init
git clone https://github.com/amirh-far/MNIST-digit-classifier.git
install python packages for this project:
cd MNIST-digit-classifier
conda install --file requirements.txt
open jupyter notebook in your project directory:
cd MNIST-digit-classifier
jupyter notebook
open classifier.ipynb
now run all cells.
Can be used to predict your images digits. Precision of this model on 10k images is 97 %.
This model is trained on 60k images and this value for precision is great.
Use this trained model as you wish.
Feel free to contribute to this mini project.
Also you can reach out to me via Gmail, LinkedIn or telegram (links available in my github profile: link).
Happy coding my friend :)