Multiple Handwritten Digit Recognition using Deep Learning ( TensorFlow - Keras )

Requirements

• TensorFlow (Keras)
• Python 3.5 +
• Numpy (+ mkl for Windows)
• PIL ( pillow )
• Opencv
• tkinter ( python GUI )

About Project

• Trained model on MNIST dataset Using CNN (Convolutional Nueral Network)

• Save model as 'mnist.h5' ( train_digit_recognizer.py )

• Using tkinter GUI make a canvas and write digit on it

• Using PIL get a copy of Handwritten Digits on the canvas and saved into '/img' as 'img_{image_number}.png'

• Also with the help of OpenCV by recognizing contours it can handle multiple digits

• Use saved model 'mnist.h5' to predict saved Handwritten Digit image from canvas

Screenshots

• Drawing canvas ...

• Output Imgae ...

• Saved Image using PIL-ImageGrab ...