This Python project Recognizes Handwritten digits using Artificial Neural Network
Requirements : Python 3.5,sklearn,opencv,numpy,imutils
train.py file is used for training the model
test.py file is used for Testing the model on test dataset,It picks up the model from model.pkl file
model.pkl file is our trained Neural Network
test_on_image.py file is used to test our ANN model on actual images
sample_images represent the images on which the model is tested
One of the Sample Image is as follows:
Recognized digits are as Follows:
Another sample image is ,
Its result is:
As we can see in above image ,one of our digit gets classified incorrectly!
One of the key thing to remember here is that extracting digit area from the natural image requires thresholding which is ofcourse in some cases leads to error,hence testing images should have proper exposure and evenly distributed brightness