An ensemble of homogeneous models for handwritten digit classification

In this example, we will use a dataset called The Street View House Numbers (SVHN) dataset from http://ufldl.stanford.edu/housenumbers/. The dataset is also provided in the GitHub in .hd5f format.

This dataset is a real-world dataset and is obtained from house numbers in Google Street View images.

We use Google Colab to train our models. In the first phase, we build a single model using Keras. In the second phase, we ensemble multiple homogeneous models and ensemble the results.

Getting ready

The dataset has 60,000 house number images. Each image is labeled between 1 and 10. Digit 1 is labelled as 1, digit 9 is labelled as 9, and digit 0 is labelled as...