Kaggle_MNIST_Keras

API Model in Keras for Kaggle MNIST competition

How to run:

1. Download train.csv and test.csv from Kaggle: (https://www.kaggle.com/c/digit-recognizer/data)

2. Move train.csv and test.csv in data directory

3. Run prepare_dataset.py in order to create the dataset

4. Run train.py to train the model if you want to see loss plot enter below command in your terminal: tensorboard --logdir logs click on the server address to see plots online in your web browser

5. Run evaluate_model_on_test_data.py in order to create submission.csv which you could use to upload it in Kaggle

*** TRAINING AND VALIDATION ACCURACIES ARE ABOVE 99%

---

Model Graph:

First Layer Filters Before Training the Model:

First Layer Filters After Training the Model:

Losses and Accuracies: (Blue ---> Validation, Orange ---> Training)