Skip to content

Latest commit

 

History

History
38 lines (27 loc) · 1.35 KB

README.md

File metadata and controls

38 lines (27 loc) · 1.35 KB

Digit-Guesser

GUI that uses AI to predict a drawn digit. The AI is a simple convolutionnal neural network (trained from scratch). The dataset was made by myself. I also used MNIST but with my dataset I got a better accuracy.

How to use

git clone :

git clone [email protected]:LucasColas/Digit-Guesser.git

Go to the directory :

cd Digit-Guesser

Install the requirements (with pip for exemple) :

pip install -r requirements.txt

and then run the code :

python main.py

Draw and predict a digit

Draw a digit by pressing the left button of your mouse and going over the gui. Unpress the left button when you don't want to draw anything.

press enter. And it will predict the digit :

Other keyboard inputs. Press the delete key to erase the current draw on the gui.

Train the neural network

Feel free to add more images. You can add more images to the dataset with images made thanks to the gui. Uncomment self.save_label(2) in gui/gui.py (line 93). The number passed in parameter in save_label is the label of the image. Then to train the neural network, run the cells of DeepLearning/deep_learning.ipynb.