usage: Segment a tomograph using convolutional neural network. Please run this program from the GUI in e2projectmanager.py.

Option Type Description
--version None show program's version number and exit
--trainset None Training set.
--from_trained str Train from an existing network
--nnet str Trained network input (nnet_save_xx.hdf)
--nettag str Tag of the output neural net file. Will use the tag of good particles in training set by default.
--learnrate float Learning rate
--niter int Training iterations
--ncopy int Number of copies for each particle
--batch int Batch size for the stochastic gradient descent. Default is 20.
--nkernel str Number of kernels for each layer, from input to output. The number of kernels in the last layer must be 1.
--ksize str Width of kernels of each layer, the numbers must be odd. Note the number of layers should be the same as the nkernel option.
--poolsz str Pooling size for each layer. Note the number of layers should be the same as the nkernel option.
--trainout None Output the result of the training set
--training None Doing training
--tomograms str Tomograms input.
--applying None Applying the neural network on tomograms
--outtag str Tag of the segmentation output. When left empty, the segmentation will be saved to 'segmentations/<tomogram name>__<neural network tag>_seg.hdf'. When set, the output will be written to 'segmentations/<tomogram name>__<outtag>.hdf'
--threads int Number of thread to use when applying neural net on test images. Not used during trainning
--ppid int Set the PID of the parent process, used for cross platform PPID
--device str For Convnet training only. Pick a device to use. chose from cpu, gpu, or gpuX (X=0,1,...) when multiple gpus are available. default is cpu