# Copyright 2019 Diamond Light Source Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
.. module:: ccpi_denoising_gpu_3D
:platform: Unix
:synopsis: GPU modules of CCPi-Regularisation Toolkit (CcpiRegulToolkitGpu)
.. moduleauthor:: Daniil Kazantsev <scientificsoftware@diamond.ac.uk>
"""
from savu.plugins.plugin import Plugin
from savu.plugins.driver.gpu_plugin import GpuPlugin
from savu.plugins.utils import register_plugin
import numpy as np
import subprocess as sp
from savu.core.iterate_plugin_group_utils import enable_iterative_loop, \
check_if_end_plugin_in_iterate_group, setup_extra_plugin_data_padding
from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, PD_TV, LLT_ROF, TGV, NDF, Diff4th
from ccpi.filters.regularisers import PatchSelect, NLTV
[docs]@register_plugin
class CcpiDenoisingGpu3d(Plugin, GpuPlugin):
def __init__(self):
super(CcpiDenoisingGpu3d, self).__init__("CcpiDenoisingGpu3d")
self.slice_dir = None
self.device = None
@setup_extra_plugin_data_padding
def set_filter_padding(self, in_pData, out_pData):
self.pad = self.parameters['padding']
pad_slice_dir = '%s.%s' % (self.slice_dir[0], self.pad)
pad_dict = {'pad_directions': [pad_slice_dir], 'pad_mode': 'edge'}
in_pData[0].padding = pad_dict
out_pData[0].padding = pad_dict
@enable_iterative_loop
def setup(self):
in_dataset, out_dataset = self.get_datasets()
pattern_type = self.parameters['pattern']
in_pData, out_pData = self.get_plugin_datasets()
in_pData[0].plugin_data_setup(pattern_type, 'single')
out_dataset[0].create_dataset(in_dataset[0])
out_pData[0].plugin_data_setup(pattern_type, 'single')
self.slice_dir = list(in_dataset[0].get_slice_dimensions())
procs = self.exp.meta_data.get("processes")
procs = len([i for i in procs if 'GPU' in i])
nSlices = int(np.ceil(in_dataset[0].get_shape()[self.slice_dir[0]] / float(procs)))
core_dims_index = list(in_dataset[0].get_core_dimensions())
core_dims_size = 1
for core_index in core_dims_index:
core_dims_size *= in_dataset[0].get_shape()[core_index]
# calculate the amount of slices than would fit the GPU memory
gpu_available_mb = self.get_gpu_memory()[0]/procs # get the free GPU memory of a first device if many
slice_size_mbbytes = int(np.ceil((core_dims_size * 1024 * 4) / (1024 ** 3)))
# calculate the GPU memory required based on 3D regularisation restrictions (avoiding CUDA-error)
if 'ROF_TV' in self.parameters['method']:
slice_size_mbbytes *= 8
if 'FGP_TV' in self.parameters['method']:
slice_size_mbbytes *= 12
if 'SB_TV' in self.parameters['method']:
slice_size_mbbytes *= 10
if 'PD_TV' in self.parameters['method']:
slice_size_mbbytes *= 8
if 'LLT_ROF' in self.parameters['method']:
slice_size_mbbytes *= 12
if 'TGV' in self.parameters['method']:
slice_size_mbbytes *= 15
if 'NDF' in self.parameters['method']:
slice_size_mbbytes *= 5
if 'Diff4th' in self.parameters['method']:
slice_size_mbbytes *= 5
if 'NLTV' in self.parameters['method']:
slice_size_mbbytes *= 8
slices_fit_total = int(gpu_available_mb / slice_size_mbbytes) - 2*self.parameters['padding']
if nSlices > slices_fit_total:
nSlices = slices_fit_total
self._set_max_frames(nSlices)
[docs] def pre_process(self):
self.device = 'gpu'
if self.parameters['method'] == 'ROF_TV':
# set parameters for the ROF-TV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'number_of_iterations': self.parameters['max_iterations'],
'time_marching_parameter': self.parameters['time_step'],
'tolerance_constant': self.parameters['tolerance_constant']}
if self.parameters['method'] == 'PD_TV':
# set parameters for the PD-TV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'number_of_iterations': self.parameters['max_iterations'],
'tolerance_constant': self.parameters['tolerance_constant'],
'methodTV': 0,
'nonneg': 0,
'PD_LipschitzConstant': 8}
if self.parameters['method'] == 'FGP_TV':
# set parameters for the FGP-TV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'number_of_iterations': self.parameters['max_iterations'],
'tolerance_constant': self.parameters['tolerance_constant'],
'methodTV': 0,
'nonneg': 0}
if self.parameters['method'] == 'SB_TV':
# set parameters for the SB-TV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'number_of_iterations': self.parameters['max_iterations'],
'tolerance_constant': self.parameters['tolerance_constant'],
'methodTV': 0}
if self.parameters['method'] == 'TGV':
# set parameters for the TGV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'alpha1': self.parameters['alpha1'],
'alpha0': self.parameters['alpha0'],
'number_of_iterations': self.parameters['max_iterations'],
'LipshitzConstant': self.parameters['lipshitz_constant'],
'tolerance_constant': self.parameters['tolerance_constant']}
if self.parameters['method'] == 'LLT_ROF':
# set parameters for the LLT-ROF method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'regularisation_parameterLLT': self.parameters['reg_parLLT'],
'number_of_iterations': self.parameters['max_iterations'],
'time_marching_parameter': self.parameters['time_step'],
'tolerance_constant': self.parameters['tolerance_constant']}
if self.parameters['method'] == 'NDF':
# set parameters for the NDF method
if self.parameters['penalty_type'] == 'huber':
# Huber function for the diffusivity
penaltyNDF = 1
if self.parameters['penalty_type'] == 'perona':
# Perona-Malik function for the diffusivity
penaltyNDF = 2
if self.parameters['penalty_type'] == 'tukey':
# Tukey Biweight function for the diffusivity
penaltyNDF = 3
if self.parameters['penalty_type'] == 'constr':
# Threshold-constrained linear diffusion
penaltyNDF = 4
if self.parameters['penalty_type'] == 'constrhuber':
# Threshold-constrained huber diffusion
penaltyNDF = 5
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'edge_parameter': self.parameters['edge_par'],
'number_of_iterations': self.parameters['max_iterations'],
'time_marching_parameter': self.parameters['time_step'],
'penalty_type': penaltyNDF,
'tolerance_constant': self.parameters['tolerance_constant']}
if self.parameters['method'] == 'Diff4th':
# set parameters for the Diff4th method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'edge_parameter': self.parameters['edge_par'],
'number_of_iterations': self.parameters['max_iterations'],
'time_marching_parameter': self.parameters['time_step'],
'tolerance_constant': self.parameters['tolerance_constant']}
if self.parameters['method'] == 'NLTV':
# set parameters for the NLTV method
self.pars = {'algorithm': self.parameters['method'],
'regularisation_parameter': self.parameters['reg_parameter'],
'edge_parameter': self.parameters['edge_par'],
'number_of_iterations': self.parameters['max_iterations']}
return self.pars
[docs] def process_frames(self, data):
input_temp = np.nan_to_num(data[0])
input_temp[input_temp > 10 ** 15] = 0.0
self.pars['input'] = input_temp
# Running Ccpi-RGLTK modules on GPU
if self.parameters['method'] == 'ROF_TV':
(im_res, infogpu) = ROF_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'],
self.parameters['GPU_index'])
if self.parameters['method'] == 'FGP_TV':
(im_res, infogpu) = FGP_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['tolerance_constant'],
self.pars['methodTV'],
self.pars['nonneg'],self.parameters['GPU_index'])
if self.parameters['method'] == 'PD_TV':
(im_res, infogpu) = PD_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['tolerance_constant'],
self.pars['methodTV'],
self.pars['nonneg'],
self.pars['PD_LipschitzConstant'],
self.parameters['GPU_index'])
if self.parameters['method'] == 'SB_TV':
(im_res, infogpu) = SB_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['tolerance_constant'],
self.pars['methodTV'], self.device)
if self.parameters['method'] == 'TGV':
(im_res, infogpu) = TGV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['alpha1'],
self.pars['alpha0'],
self.pars['number_of_iterations'],
self.pars['LipshitzConstant'],
self.pars['tolerance_constant'], self.device)
if self.parameters['method'] == 'LLT_ROF':
(im_res, infogpu) = LLT_ROF(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['regularisation_parameterLLT'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'], self.device)
if self.parameters['method'] == 'NDF':
(im_res, infogpu) = NDF(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['edge_parameter'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['penalty_type'],
self.pars['tolerance_constant'], self.device)
if self.parameters['method'] == 'DIFF4th':
(im_res, infogpu) = Diff4th(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['edge_parameter'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'], self.device)
if self.parameters['method'] == 'NLTV':
pars_NLTV = {'algorithm': PatchSelect,
'input': self.pars['input'],
'searchwindow': 9,
'patchwindow': 2,
'neighbours': 17,
'edge_parameter': self.pars['edge_parameter']}
H_i, H_j, Weights = PatchSelect(pars_NLTV['input'],
pars_NLTV['searchwindow'],
pars_NLTV['patchwindow'],
pars_NLTV['neighbours'],
pars_NLTV['edge_parameter'], self.device)
parsNLTV_init = {'algorithm': NLTV,
'input': pars_NLTV['input'],
'H_i': H_i,
'H_j': H_j,
'H_k': 0,
'Weights': Weights,
'regularisation_parameter': self.pars['regularisation_parameter'],
'iterations': self.pars['number_of_iterations']}
im_res = NLTV(parsNLTV_init['input'],
parsNLTV_init['H_i'],
parsNLTV_init['H_j'],
parsNLTV_init['H_k'],
parsNLTV_init['Weights'],
parsNLTV_init['regularisation_parameter'],
parsNLTV_init['iterations'])
del H_i, H_j, Weights
return im_res
def _set_max_frames(self, frames):
self._max_frames = frames
[docs] def get_gpu_memory(self):
command = "nvidia-smi --query-gpu=memory.free --format=csv"
memory_free_info = sp.check_output(command.split()).decode('ascii').split('\n')[:-1][1:]
memory_free_values = [int(x.split()[0]) for i, x in enumerate(memory_free_info)]
return memory_free_values
def nOutput_datasets(self):
return 1
# total number of output datasets
[docs] def nOutput_datasets(self):
if check_if_end_plugin_in_iterate_group(self.exp):
return 2
else:
return 1
# total number of output datasets that are clones
[docs] def nClone_datasets(self):
if check_if_end_plugin_in_iterate_group(self.exp):
return 1
else:
return 0