https://github.com/isosnovik/nets

Draw your net
https://github.com/isosnovik/nets

architecture deep-learning draw keras neural-network segmentation visualization

Last synced: 4 days ago
JSON representation

Draw your net

• Host: GitHub
• URL: https://github.com/isosnovik/nets
• Owner: ISosnovik
• Created: 2016-12-26T17:05:46.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2017-09-01T19:22:40.000Z (about 7 years ago)
• Last Synced: 2024-08-04T03:11:09.519Z (3 months ago)
• Topics: architecture, deep-learning, draw, keras, neural-network, segmentation, visualization
• Language: Python
• Homepage:
• Size: 56.6 KB
• Stars: 48
• Watchers: 4
• Forks: 10
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Nets

It is an easy-to-use framework for drawing your favorite network's architecture.

It is inspired by the juicy schemas from [SegNet](http://mi.eng.cam.ac.uk/projects/segnet/)

## Usage

The API is close to this of [Keras](https://keras.io/getting-started/sequential-model-guide/)

You don't have to think about filter size, strides and activation function. It is all about shapes, colors and fonts.

```python

from nets.PlaneLayer import *

from nets.neural_net import NN

nn = NN()

# Encoder

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Pooling(pool_size=2))

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Pooling(pool_size=2))

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Pooling(pool_size=2))

# Decoder

nn.add(Upsampling(factor=2))

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Upsampling(factor=2))

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Upsampling(factor=2))

nn.add(Conv2D())

nn.add(Conv2D())

nn.add(Conv2D())

```

To draw the net, the architecture should be processed

```python

nn.compile()

nn.add_legend(bbox_to_anchor=(0.2, -0.1, 0.6, 0.1), fontsize=26)

nn.draw()

```

Pretty simple. That's it

![img](src/net.png)