Build and train a deep neural network model, to recognize and classify images of clothing, like sneakers and shirts in the Fashion-MNIST dataset, using Keras and Tensorflow.
Here I add a validation set to test for overfitting while training.
During training I looked at the training and validation losses, there was overfitting. So I applied two ways to prevent my model from overfitting, Early Stopping and Dropout.
After the model is trained, I used it to perform inference on 30 images and print the labels in green if my model's prediction matches the true label. On the other hand, if my model's prediction doesn't match the true label, I print the label in red.
This is a mini project I did during the Machine Learning nanodegree at Udacity.