Mainly deprecated since Cleverhans implemented model zoo.

I still use it for experimentation and might roll some new features.

The library was created in order to extend Cleverhans models. The main goal is to ease the design process for new models by introducing the train_graph.py abstraction and other layers such as lrn.py.

The inspiration comes from Cleverhans. For the moment all models are extended to use name_scopes. Moreover, new architectures will be implemented in the get_model.py file.

Installation

pip Installation:

pip install pip install dnnmodels

Manual Installation

git clone https://github.com/NullConvergence/dnn-models pip install -e ./dnn-models

Examples

• At the moment there is only one example: mnist.py