Introduction to IEEE STANDARDS and its different types.pptx
Digit recognizer
1. Digit-Recognizer
… a wonderful application of CNN
Made by: Awismrit Parida
(Eckovation)
Course: Machine Learning in Python
2. Abstract
Recognizes real time hand written digits using CNN (Convolution
neural networks)
Repository Link: https://github.com/AWIS99/Digit-Recognizer
3. Requirements
Python v3.7
TensorFlow v2.1.0
OpenCV v3.4.7
Keras
NumPy
*one can install anaconda which solves most of the dependencies
4. Steps of Development
Make sure that your development environment is ready and fulfils all
the previously-mentioned requirements.
Download the dataset Fashion MNIST which is built into the
TensorFlow and split it into training data, training labels testing data
and testing labels.
6. Define the model (CNN) according to the given architecture below in
this report, followed by compilation, training and evaluation.
7. Save the generated model for later use.
Next is OpenCV part. Import all required libraries. Load the saved
model and execute the OpenCV code ( get the code from repository ).
It’s done.
8. Usage
Download all the files in the repository.
Copy the file path of 'digit.h5' which is the trained model with a test
accuracy of over 99%.
Open the 'OpenCV_code.py' file in any text editor and replace the
path inside the load_model function and save it.
Now run the file in your configured terminal/cmd.
Take any blue coloured pointer and start drawing in front of webcam.
The detected digits will be shown in the bottom left corner.
Enjoy!!
You can also do the same in Jupiter notebook with the ipynb files.