Veröffentlichte Applikation

Use Pipenv:

pipenv install

Use pip:

pip install -r requirements.txt

A jupyter notebook for training a convolutional neural network (CNN) on the MNIST dataset

• Accuaracy: 99% on random test data from the MNIST dataset
• A visualisation of the Loss-Function over the training process
• Testplots
• Autodownloader and split for the training and validation data

Modell-Architecture:

Image Source

Framerwork: pytorch

A jupyter notebook for training a depp neural network (DNN) on the MNIST dataset

• Accuaracy: 98% on random test data from the MNIST dataset
• A visualisation of the Loss-Function over the training process
• Testplots
• Autodownloader and split for the training and validation data

Modell-Architecture:

Image Source

Framerwork: pytorch

A streamlit Application for Testing the CNN and the DNN

Start the Application:

streamlit run app.py

Draw your own numbers inside a canvas and let the CNN take a guess.

Compare the Results from the DNN with the CNN

Framerwork: streamlit

Thanks to Andreas Weber for the canvas idea