DnnDraw

DnnDraw is a framework for visualizing neural networks via Python programming.

How to install

apt-get install graphviz graphviz-dev

Install from pypi:

pip install dnndraw

Install from source:

pip install git+https://github.com/0400H/DnnDraw.git

---

Examples

Tinydnn

import dnndraw

dnn = dnndraw.graph(name="tinydnn")

# first layer
dnn.add_node(in_nodes=[], node_info={'name': 'layer_1', 'Type': 'Conv3D', 'kernel': [1, 1, 1], 'stride': [1, 1, 1], 'padding': 'none', 'normal, relu': 'True'})

dnn.add_node(in_nodes=['layer_1'], node_info={'name': 'layer_2', 'Type': 'Conv3D', 'kernel': [3, 3, 3], 'stride': [1, 1, 1], 'padding': 'none', 'normal, relu': 'True'})

dnn.add_node(in_nodes=['layer_1'], node_info={'name': 'layer_3', 'Type': 'Conv3D', 'kernel': [3, 3, 3], 'stride': [1, 1, 1], 'padding': 'none', 'normal, relu': 'True'})

# end layer
dnn.add_node(in_nodes=['layer_2', 'layer_3'], node_info={'name': 'layer_4', 'Type': 'Concat'})

print(dnn.source())
dnn.export(format='png') # format: png, svg, pdf, ...
dnn.show()

---

Dev or Build Example

• via develop mode

  python3 ./setup.py develop
  cd example
  python ./tinydnn.py
  

• via editable mode

  pip install -e .
  cd example
  python ./tinydnn.py
  

• via PYTHONPATH

  export PYTHONPATH=`pwd`
  cd example
  python ./tinydnn.py