A Convolutional Neural Network designed to classify digits. CNN is trained and tested on famous MNIST Digit Dataset. This Convolution Neural Network gains an accuracy of 0.9996 on Training dataset and accuracy of 0.99428 on Test dataset. MNIST Digit image dataset is obtained from Kaggle website.
Details of Files:
- train.csv consists of labeled training data.
- test.csv consists of test data.
- DigitRecognizer.py consists of Digit Recognizer CNN Model.
- digit_recognizer_model.h5 consists of trained model weights.
- Result.csv consists of output of CNN tested on test data.
CNN is built using following libraries: Pandas, numpy, sklearn and keras.