Digit Recognizer
Python Jupyter Notebook with Convolutional Neural Network digit recognizer implemented in Keras. It's Google Colab ready.
Part of the Kaggle competition.
Submitted Kernel with 0.995 score.
Check out corresponding Medium article:
Digit Recognizer - Introduction to Kaggle Competitions with Image Classification Task (0.995)
Data
Dataset: MNIST Handwritten digits
Description: Classification of handwritten digits, 10 classes (0-9).
Training: 37.8k (0.9) images
Validation: 4.2k (0.1) images
Testing: 28k images
Model
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_22 (Conv2D) (None, 28, 28, 32) 832
_________________________________________________________________
conv2d_23 (Conv2D) (None, 28, 28, 32) 25632
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, 14, 14, 32) 0
_________________________________________________________________
dropout_7 (Dropout) (None, 14, 14, 32) 0
_________________________________________________________________
conv2d_24 (Conv2D) (None, 14, 14, 64) 18496
_________________________________________________________________
conv2d_25 (Conv2D) (None, 14, 14, 64) 36928
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, 7, 7, 64) 0
_________________________________________________________________
dropout_8 (Dropout) (None, 7, 7, 64) 0
_________________________________________________________________
flatten_4 (Flatten) (None, 3136) 0
_________________________________________________________________
dense_8 (Dense) (None, 8192) 25698304
_________________________________________________________________
dropout_9 (Dropout) (None, 8192) 0
_________________________________________________________________
dense_9 (Dense) (None, 2048) 16779264
_________________________________________________________________
dropout_10 (Dropout) (None, 2048) 0
_________________________________________________________________
dense_10 (Dense) (None, 10) 20490
=================================================================
Total params: 42,579,946
Trainable params: 42,579,946
Non-trainable params: 0
_________________________________________________________________
Training
Results
Kaggle score: 0.995
Author
Greg (Grzegorz) Surma