DeepGUI is a graphical user interface which generates Deep Learning Frameworks codes for you. You can just add, remove, and edit layers in a graphical way and the interface generates python code for you. Cool, ha?
DeepGUI is built using Electron.js framework.
The new features are as follows:
-
Now you can use transfer learning.
-
The bug in the learning rate is fixed.
What's New!? Now you can ...
-
Generate code using PyTorch framework (although, because PyTorch sequential models doesn't support time analysis, recurrent layers are not supported).
-
Add new layers between other layers. Previously it was only possible to add layers at the end of the sequence.
-
Save the diagram into a .dgui file and load it back later.
The specifications of this version:
-
These layers are available in this version:
- Dense,
- Convolution (1D, 2D, and 3D) layers,
- Pooling (Max and Average) (1D, 2D, and 3D)
- RNN, LSTM, and GRU layers,
- Embedding layer,
- Batch Normalization layer,
- Flatten layer,
- Dropout layer,
- and Activation layer.
-
Only TensorFlow framework is accessible. PyTorch will be added soon.
-
Basic configurations of each layer is added. Advanced configurations will be added soon.
-
Only Sequential models are available in this version.
I provide some screenshots of the GUI here. I know that this configuration is not OK for a 10-class image dataset classification. It's just a simple example.
