This repository has been archived by the owner on May 13, 2019. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathChangelog.txt
271 lines (172 loc) · 11 KB
/
Changelog.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
================ Release 0.1.4 ================
Major notes:
________________________________________________________________
Patch from Kent Williams, fixes pthreads issue on OSX as well as ITK
library issue for ITK 4.0. Thanks Kent!
Changed gradient computation for relaxation back to use original
jacobian determinant calculation instead of new adjoint formulation.
New formulation was causing pinching of ventricles during brain atlas
generation, more testing is needed.
Added momemtum optimization for LDMM deformation estimation based on
Vialard et. al.'s "Diffeomorphic 3D Image Registration via Geodesic
Shooting Using an Efficient Adjoint Calculation". Atlas computation
not added yet.
Added support for CUDA 4.0
Added additional options for saving results, including per-deformation
energy files during atlas building.
Added *some* support for vtk files to txApply.
Added AutoStepReduce parameter, reduces step size automatically for
non-converging deformations.
Fixed voxel-wise median computation for distributed atlas building.
Details:
________________________________________________________________
2011-09-09 r1458 Finished reverting LDMM relaxation code back to old (jacobian determinant based) algorithm. This fixed problem with pinched ventricles during atlas building. New adjoint-based algorithm still included in LDMMIterator[CPU/GPU].cxx, but code must be edited and recompiled to select it.
2011-07-27 r1362 Modification to use original gradient calculation (instead of _shared version), causes problems on some GPUs. Cause is still unclear.
2011-06-15 r1296 Added option to turn off scaling by jacobian determinant during atlas construction
2011-06-14 r1295 Quick addition to run adjoint shooting optimization using LDMMMultiWarp
2011-06-10 r1293 Fixed faster version (at the expense of memory) of GPU LDMMAdShootingWarp. New implementation is now the default, can be changed via preprocessor definition in LDMMAdShootingIteratorGPU.h
2011-06-10 r1291 Modifications to fix non-CUDA builds
2011-06-09 r1289 GPU version of adjoint shooting matches CPU version, GPU test added to ctest and passes.
2011-06-07 r1286 Added debugging info to catch NaNs in individual deformations
2011-06-05 r1282 Merged adjoint shooting optimization code from Nikhil's branch into trunk and added testing. All tests pass.
2011-06-01 r1279 Additions necessary for CUDA 4.0
2011-05-24 r1275 Added option to reparameterize deformations to have constant speed, but this doesn't seem to have much effect on the result.
2011-05-17 r1272 Added options for whether to save / write vector fields, deformed images, etc.
2011-04-29 r1270 fixed writing out of individual energy files
2011-04-25 r1269 Modification to txApply to properly handle vtk surfaces
2011-03-23 r1252 Fixed voxelwise median and PCA bug for multi-gpu systems
2011-03-21 r1251 Added AutoStepReduce param, which reduces the step size for a deformation whenever an increase in energy is detected. Along with TrimmedMeanSize, allows atlases with outliers to be stably calculated.
2011-03-21 r1250 Integrated new energy history classes, added option to save energy in XML file for LDMM/Greedy warp/atlas applications
2011-03-17 r1248 distributed voxelwise median calculation works for CPU, but getting deformed images in main thread from mutliple GPUs is still broken
================ Release 0.1.3 ================
Major notes:
________________________________________________________________
Added option to save energy in XML format
Added MultiGaussian Kernel as option instead of DiffOper
Added Trimmed Mean option for atlas building
Added option to use voxel-wise mean instead of median when computing atlas
Bug for inverting affine transforms in txWerks fixed
Details:
________________________________________________________________
2011-03-17 r1247 Change to copyright holders list
2011-03-14 r1246 Added TrimmedMeanSize option to LDMM Atlas Building. When TrimmedMeanSize > 0, only the TrimmedMeanSize images 'closest' to the mean will be used in mean calculation.
2011-03-08 r1245 added option to do voxel-wise median computation instead of mean
2011-03-04 r1241 merged updates to PCA computation from LowSparse branch to trunk
2011-02-28 r1236 small fix to make LDMMIteratorGPU work correctly when USE_LV is false
2011-02-22 r1235 added the option to save energy history as a file to LDMMWarp applications -- especially important for LDMMMultiWarp application, since output from multiple nodes gets jumbled together.
2011-02-18 r1233 Modifications to power method to run correctly for mpi jobs
2011-02-17 r1231 building distributed greedy atlas application was accidentally omitted, this has been added
2011-02-16 r1230 Changed PCA to use XX^T trick, not much faster but uses a lot less memory
2011-02-16 r1229 added option to do PCA step during greedy atlas building
2011-02-11 r1228 Modified LDMM applications to optionally use MultiGaussianKernel instead of DiffOper
2011-02-04 r1220 added fix for complex with swig
2011-02-03 r1218 Changed Complex type defined in CUDACommon to cplComplex to avoid name collision
2011-02-03 r1216 Change to fix finding cudpp lib for cuda
2011-01-26 r1206 fixed bug for inverting affine transform txWerks
================ Release 0.1.2 ================
Major notes:
________________________________________________________________
Major bugfix to differential operator for non-unit-spacing images,
this affects all test results.
Naming changes to warp output files: names now contains both the
static and moving image name. This causes long names, but avoids
ambiguity or accidental overwriting of output.
Naming changes to parameters: Atlas output parameters now use
directions DefToMean or DefMean instead of the jumble of 0To1 or other
naming conventions, hopefully making the parameters more
comprehensible. Output filename suffixes also changed to match.
Cuda 3.2 is now supported (along with 3.1)
Error message has been added when running GPU applications on hardware
that doesn not support the necessary compute capability level
Details:
________________________________________________________________
2011-01-10 r1203 Fixed a few problems with BuildAtlasWerks.sh script
2010-12-30 r1199 added check for required CUDA compute capability for
LDMM applications
2010-12-30 r1198 Fixes to LDMMVelShootingIterator to use correct
spacing (thanks Linh!), as well as fixing some settings for upsampling
vector fields. Updated testing.
2010-12-15 r1197 Fixed a bug in LDMMVelShootingIterator.cxx
2010-12-14 r1196 Fix for txApply affine deformation, wasn't using
correct origin for nonzero image origins
2010-12-14 r1195 Fix for CUDA 3.1/3.2 incompatibilities
2010-12-13 r1194 Bugfix to differential operator. Caused problems for
non-unit spacing, especially non-integer non-unit spacing. This
breaks some testing, will fix that soon.
2010-12-13 r1192 Initial check-in of code and testing for LDMM
velocity shooting optimization. There may still be bugs, and the
testing is not added to the automated testing yet (it doesn't pass)
2010-12-09 r1191 Make change to compute with CUDA 3.2
2010-12-08 r1188 Name changes for testing code -- reflects change to
warp output including names of static and moving image in output image
by default
2010-12-08 r1187 Changes to DiffOper -- a double-precision
differential operator can now be created as DiffOperT<double>,
DiffOper is now a typedef for DiffOperT<float>
2010-12-02 r1184 A few additions to python wrapping (added a 'clone'
function to datatypes), and a bugfix allowing CPU atlas building to
write out Alpha images.
2010-12-01 r1182 Added better jacobian computation to velocity
shooting. Doesn't match momentum shooting exactly, but quite close.
2010-11-30 r1180 Fixed bug in alpha computation for CPU version of
LDMM relaxation, added initial version of LDMM velocity shooting (CPU
only at this point, and still need to have better calculation of
jacobian)
2010-11-23 r1178 Added 'levels' to parameters, so that debug
parameters can be added without needing to be displayed. Levels are
PARAM_REQUIRED, PARAM_COMMON, PARAM_RARE, and PARAM_DEBUG. Also fixed
bug causing Alpha0s not to be computed when only WriteAlpha0s
specified (not WriteAlphas) for LDMMWarp / LDMMMultiWarp, and modified
warp outputs to contain the moving and static image names in the
output name.
2010-11-17 r1176 Modified the parameter files to have (hopefully) more
comprehensible names. Backwards compatibility for parameter files is
preserved (the old parameter names are still accepted), but HFields
generated by atlas building programs now have new filenames. Also
increased the default 'thresholdDistance' in
HField3DUtils::inverseOfPoint(...) -- not sure why this threshold is
there, but it causes problems for large deformaitons.
2010-11-10 r1173 Added working MaxPert param to LDMMWarp / LDMMAtlas
(for relaxation only, shooting optimization not implemented yet). To
use it, set UseAdaptiveStepSize to 'true' and set MaxPert
appropriately (between 0 and 1, probably closer to zero)
2010-11-08 r1172 fixed a bug in AtlasWerks testing and added a trilerp
function wrapper for python access
2010-11-05 r1171 Added velocity geodesic shooting and fixed a bug in
MultiscaleManager
2010-11-05 r1170 Added loading of velocity fields for relaxation
optimization initialization, and fixed writing of alpha0 images and
velocity fields
2010-11-03 r1169 Modifications to add checkpoint restart to
LDMMShootingGPU Atlas Building -- still need to implement that for
other apps
2010-10-26 r1166 Added Code/Python
2010-10-19 r1161 Update FFTSolver3D functions with streaming form
2010-10-18 r1160 Fixed bug in the composeHV*_tex function
2010-10-18 r1159 Add resample function using texture that is faster
================ Release 0.1.1 ================
Major Notes:
________________________________________________________________
Global name change from CompOnc to AtlasWerks
Added build script
Details
________________________________________________________________
2010-10-14 r1155 fixing problems with python wrapping
2010-10-14 r1154 adding build script for release tarballs
2010-10-12 r1152 added GPU version of shooting. Not all that much
faster than CPU version, mostly just for testing
2010-10-08 r1151 Modification to GeodesicShooting to use adjoint
equations for evolution and a bugfix to shooting atlas building
2010-10-08 r1150 Global namechange from CompOnc to AtlasWerks
2010-10-06 r1148 Modified test images to have non-unit non-uniform
spacing, non-uniform size, and nonzero origin to catch more problems
during regression. This uncovered a few problems (specifically
AtlasWerks handling of spacing on CPU vs. GPU and issues with non-unit
spacing in shooting optimization), but testing parameters were
adjusted to allow these to pass for now (since they were passing
before). These will have to be addressed later (or perhaps not in
AtlasWerks since it's being superceded by GreedyAtlas anyway).
2010-10-05 r1147 fixing compilation errors for non-GPU code
2010-10-05 r1145 Fixed bug writing out forward hfield when inverse
hfield was requested
================ Release 0.1.0 ================