Here are test datasets, deep learning models, and binary executable (Linux version Vaa3D 3.433) for the DeepNeuron testing.
The paper citation is the following:
Zhi Zhou, Hsien-Chi Kuo, Hanchuan Peng, Fuhui Long, "DeepNeuron: An Open Deep Learning Toolbox for Neuron Tracing," doi: https://doi.org/10.1101/254318
It can be downloaded at https://www.biorxiv.org/content/early/2018/01/25/254318.full.pdf
untar "Vaa3D_CentOS_64bit_DeepNeuron.tar.gz" to obtain the binary on Linux (preferably CentOS)
unzip "testing_data.zip" to obtain all testing dataset.
unzip "deep_learning_models.zip" to obtain all needed deep learning models.
Run the "start_vaa3d.sh" in the folder "Vaa3D_CentOS_64bit_DeepNeuron". Then go to main menu and choose "Plug-In", and then click "DeepNeuron_toolbox" menu. You will find five modules:
- Neurite Signal Detection: use deep learning models under "deep_learning_models/2_labels/2D_CNN_5_layers" folder.
- Neurite Connection: use deep learning models under "deep_learning_models/2_labels/siamese_networks" folder.
- Smart Pruning: use deep learning models under "deep_learning_models/2_labels/2D_CNN_5_layers" folder.
- Reconstruction Evaluation: use deep learning models under "deep_learning_models/2_labels/2D_CNN_5_layers" folder.
- Classification of Dendrites and Axons: use deep learning models under "deep_learning_models/3_labels" folder. You can choose "2D_CNN_5_layers" or "2D_CNN_6_layers".
Note that you will need to open an image stack (in V3DPBD format) from "testing_data" folder in Vaa3D before running any DeepNeuron modules. In "Neurite Connection", "Smart Pruning", and "Reconstruction Evaluation" modules, you need to load a reconstruction file (in SWC format), which can be obtained from "testing_data" folder or generated by "Neurite Signal Detection" module. You can also load these SWC files into Vaa3D to visualize or quantify or analyze.
Questions? Solve the issues at the Vaa3D NITRC forum https://www.nitrc.org/forum/forum.php?forum_id=1553 .