added niftyreg resampling for linear and cubic B-splines. Fixed compi…

…lations issue (tested on Linux gcc 5.4 cuda 10.2). Resampling and Gradient are supported on cpu and gpu (but no gradient wrt the warped image itself)

niftynet/contrib/layer/resampler_optional_niftyreg.py

@@ -0,0 +1,32 @@
+# Flag stating whether C++/CUDA image resampling is available
+HAS_NIFTYREG_RESAMPLING = False
+
+try:
+    from niftyreg_image_resampling import NiftyregImageResamplingLayer
+    import niftyreg_image_resampling as resampler_module
+
+    ResamplerOptionalNiftyRegLayer = NiftyregImageResamplingLayer
+
+    HAS_NIFTYREG_RESAMPLING = True
+except ImportError:
+    import tensorflow as tf
+
+    tf.logging.warning('''
+    niftyreg_image_resampling is not installed; falling back onto
+    niftynet.layer.resampler.ResamplerLayer. To install
+    niftyreg_image_resampling please see
+    niftynet/contrib/niftyreg_image_resampling/README.md
+    ''')
+
+    from niftynet.layer.resampler import ResamplerLayer
+    import niftynet.layer.resampler as resampler_module
+
+    ResamplerOptionalNiftyRegLayer = ResamplerLayer
+
+
+# Passthrough of supported boundary types
+SUPPORTED_BOUNDARY = resampler_module.SUPPORTED_BOUNDARY
+
+
+# Passthrough of supported interpolation types
+SUPPORTED_INTERPOLATION = resampler_module.SUPPORTED_INTERPOLATION

niftynet/contrib/niftyreg_image_resampling/.gitignore

@@ -0,0 +1 @@
+*~

niftynet/contrib/niftyreg_image_resampling/LICENSE

@@ -0,0 +1,28 @@
+Copyright (c) 2009, University College London, United-Kingdom
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+Redistributions of source code must retain the above copyright notice,
+this list of conditions and the following disclaimer.
+Redistributions in binary form must reproduce the above copyright notice,
+this list of conditions and the following disclaimer in the documentation
+and/or other materials provided with the distribution.
+
+Neither the name of the University College London nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+THE POSSIBILITY OF SUCH DAMAGE.
+

niftynet/contrib/niftyreg_image_resampling/README.md

@@ -0,0 +1,25 @@
+# NiftyNet GPU Image Resampling Module
+
+## Purpose and Scope
+
+This module provides a faster implementation of image resampling. For most usage scenarios, it is a drop-in replacement for niftynet.layer.resampler.ResamplerLayer, however, its feature set is limited to:
+
+* ZERO (zero-padding), REPLICATE (clamping of intensities at edges), and SYMMETRIC (mirroring) boundaries
+* NEAREST (constant), LINEAR, and BSPLINE (cubic spline) interpolation.
+* Differentiation with respect to the floating image is a CPU-only operation.
+
+To provide compatibility where this module is not installed, the following module can be used: niftynet.contrib.layer.resampler_optional_niftyreg.ResamplerOptionalNiftyRegLayer. This module will try to load NiftyregImageResamplingLayer, if that fails, it defaults to niftynet.layer.resampler.ResamplerLayer.
+
+## Building and Installing
+
+Building and installing is performed as usual via setup.py.
+
+Building the module requires that a CUDA toolkit and CMake be installed, and nvcc and cmake can be found on the executables search path.
+CMake variables can be overriden through the `override` command and `--settings`/`-s` switch as a list of colon (':') separated variable-value pairs. E.g., `python setup.py override -s "CMAKE_CXX_COMPILER:/usr/bin/g++-6:CMAKE_C_COMPILER:/usr/bin/gcc-6" build`.
+
+Building was only tested with CUDA 9.0/gcc 5.4 and gcc 6.4, on Ubuntu 16.04 and 18.04.
+
+## Acknowledgements
+
+The image resampling code contained in this module is heavily based on code extracted from NiftyReg (https://sourceforge.net/projects/niftyreg/).
+

niftynet/contrib/niftyreg_image_resampling/niftyreg_gpu_resampler/CMakeLists.txt

@@ -0,0 +1,64 @@
+project(NiftyNet_gpu_resampling)
+cmake_minimum_required(VERSION 3.5)
+
+set(CMAKE_MODULE_PATH "${CMAKE_SOURCE_DIR}/cmake/modules;${CMAKE_MODULE_PATH}")
+
+set(GPU_RESAMPLING_CONFIGFILE_DIR ${CMAKE_CURRENT_BINARY_DIR} CACHE STRING "Destination directory for configured files.")
+
+add_definitions("-DGOOGLE_CUDA")
+set(GPU_RESAMPLING_LIB_TARGET niftyreg_image_resampling_ops)
+set(CMAKE_CXX_STANDARD 11)
+# This may not be portable (almost certainly isn't)!
+# -DNDEBUG: a constexpr applied to function returning std::string is causing the build to fail; the code appears to be inside an assert, so disabling asserts "solves" the problem.
+set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -std=c++11 --expt-relaxed-constexpr -DNDEBUG --disable-warnings")
+
+find_package(CUDA)
+find_package(Tensorflow REQUIRED)
+
+include_directories(nifti)
+include_directories(SYSTEM "${Tensorflow_INCLUDE_DIRS}")
+link_directories("${Tensorflow_LIBRARY_DIRS}")
+set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} ${Tensorflow_CFLAGS}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${Tensorflow_CFLAGS}")
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${Tensorflow_CFLAGS}")
+set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS}")
+
+set(GPU_RESAMPLING_CXX_SRC
+  _reg_resampling.cpp
+  _reg_tools.cpp
+  _reg_maths.cpp
+
+  niftyreg_cpu_resample_op.cpp
+  niftyreg_cpu_resample_gradient_op.cpp
+  niftyreg_cpu_resample_image_gradient_op.cpp
+
+  nifti/nifti1_io.c
+  nifti/znzlib.c
+  )
+
+if (CUDA_FOUND)
+  set(GPU_RESAMPLING_CU_SRC
+    resampleKernel.cu
+    _reg_common_cuda.cu
+    _reg_resampling_gpu.cu
+
+    niftyreg_gpu_resample_op.cu
+    niftyreg_gpu_resample_gradient_op.cu
+    )
+
+  cuda_add_library(${GPU_RESAMPLING_LIB_TARGET} SHARED
+    ${GPU_RESAMPLING_CXX_SRC}
+    ${GPU_RESAMPLING_CU_SRC}
+    )
+else ()
+  message(WARNING "No CUDA toolkit was found, if your tensorflow install was built with CUDA support, the build will fail.")
+  add_library(${GPU_RESAMPLING_LIB_TARGET} SHARED
+    ${GPU_RESAMPLING_CXX_SRC}
+    )
+endif (CUDA_FOUND)
+
+set_target_properties(${GPU_RESAMPLING_LIB_TARGET} PROPERTIES PREFIX "")
+target_link_libraries(${GPU_RESAMPLING_LIB_TARGET}
+  ${Tensorflow_LIBRARIES}
+  )
+file(GENERATE OUTPUT "${GPU_RESAMPLING_CONFIGFILE_DIR}/niftyreg_module_loader.py" INPUT "${CMAKE_CURRENT_SOURCE_DIR}/../niftyreg_module_loader.py.in")