Interactive ML webapp using Gradio and Tensorflow to identify handwritten digits using a model trained on the MNIST dataset.
-
Python 3.9 & up
-
Open Command prompt and type:
pip install gradio --user pip install tensorflow --user
-
Open Powershell in the local repository folder
-
Type:
python main.py
-
The web interface will be available at the url generated: http://localhost:7860
-
The program is also available in Google Colab which can be accessed here.
-
Copy the notebook to your Google Drive and follow the instructions to run the code.
-
Draw different digits using mouse cursor onto the sketch area and click submit.
-
A list of guesses along with how confident the trained model is for other guesses will be provided with varying percentages.
-
Click clear to refresh the sketch area and start drawing again.