Keras model about the shvn problem, with this dataset: https://www.kaggle.com/olgabelitskaya/svhn-preproccessed-fragments/data.
I improve this kernel: https://www.kaggle.com/olgabelitskaya/svhn-digit-recognition and I apply a similar architecture proposed in the official paper: https://arxiv.org/pdf/1312.6082.pdf
STEPS
- Download the dataset from https://www.kaggle.com/olgabelitskaya/svhn-preproccessed-fragments/data, and put it in the
housenumbers
directory (create it if you don't have it). - Install the dependencies with pip3 if you use PyPi with
sudo pip3 install pandas matplotlib tensorflow keras numpy ipython
- Execute the notebook with
ipython3 notebook svhnModel.ipynb