-
Notifications
You must be signed in to change notification settings - Fork 1
/
PointCloud.m
327 lines (258 loc) · 10.6 KB
/
PointCloud.m
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
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
classdef (Abstract) PointCloud < handle
methods (Abstract)
generate_global_proposals(self, num_random_proposals);
end
properties (Hidden)
pairwise_cost;
max_neighbors;
end
properties (SetAccess = protected)
points;
dimensions;
neighborhood;
num_points;
end
properties (Access = public)
assignments;
data_weight;
tol;
% For plot function
tangent_width;
tangent_thickness;
tangent_color;
point_size;
point_color;
input_point_size;
input_point_color;
verbose;
cost_function;
eps;
end
methods
function self = PointCloud(points, neighborhood, data_weight, tol)
self.points = points;
self.num_points = size(self.points,2);
self.assignments = [points; -sum(points.*points)];
self.neighborhood = neighborhood;
self.data_weight = data_weight;
self.tol = tol;
% Defaults
self.tangent_width = 0.075;
self.tangent_thickness = 1;
self.tangent_color = [0 0 0];
self.point_size = 1;
self.point_color = [0 0 1];
self.input_point_size = 1;
self.input_point_color = [1 0 0];
self.cost_function = 'default';
self.eps = 1e-7;
self.verbose = false;
% Calculate most neighbors any point
self.max_neighbors = 0;
for i = 1:self.num_points
self.max_neighbors = max(sum(neighborhood.ind1 == i), self.max_neighbors);
self.max_neighbors = max(sum(neighborhood.ind2 == i), self.max_neighbors);
end
% Add path to include
addpath(fileparts(mfilename('fullpath')));
addpath([fileparts(mfilename('fullpath')),filesep,'include']);
%% Compile if need be
compile_script('energy');
compile_script('local_optimization');
% Compile if need be
sources = {['QPBO-v1.3.src' filesep 'QPBO.cpp'], ...
['QPBO-v1.3.src' filesep 'QPBO_extra.cpp'], ...
['QPBO-v1.3.src' filesep 'QPBO_maxflow.cpp'], ...
['QPBO-v1.3.src' filesep 'QPBO_postprocessing.cpp']};
compile_script('fusion', sources);
end
% Plot functions
function plot_energy(self)
[E,U,B] = self.energy;
title(sprintf('Total cost: %2.3f -- data: %2.3f, regularization: %2.3f', E,U,B));
end
% Fuse one proposal
function fused_variables = binary_fusion(self, proposal)
% If the proposals is only one plane
if (size(proposal,2) == 1)
proposal = repmat(proposal, [1 self.num_points]);
end
if (all(size(proposal) ~= size(self.assignments)))
error('Proposal should be same size as assignment, or only one tangent plane');
end
improve = false;
[S, E, LB, num_unlabelled] = fusion_mex(...
self.cost_function, ...
self.dimensions, ...
self.assignments, ...
proposal, ...
self.points, ...
self.connectivity - 1, ...
self.neighborhood.value, ...
self.data_weight, ...
self.tol, ...
self.eps, ...
self.verbose, ...
improve);
fused_variables = sum(S);
if (self.verbose)
fprintf('Fusion info -- fused %d variables, energy: %g lb: %g unlabeled: %d.\n', ...
fused_variables, E, LB, num_unlabelled);
self.plot();
drawnow();
end
% Update
self.assignments(:,S==1) = proposal(:,S==1);
end
% fuse in random order
function batch_binary_fusion(self, proposals)
if (~isa(proposals,'cell'))
error('Proposals must be a cell array');
end
for i = 1:randperm(numel(proposals))
self.binary_fusion(proposals{i});
end
end
function fuse_until_convergence(self, proposals)
if (~isa(proposals,'cell'))
error('Proposals must be a cell array');
end
fused = 1;
while (fused > 0)
fused = 0;
for i = 1:randperm(numel(proposals))
fused = fused + self.binary_fusion(proposals{i});
end
end
end
function set.cost_function(self, cost_function)
switch(cost_function)
case {'linear','quadratic', 'default', 'length'}
self.cost_function = cost_function;
otherwise
error('Regularization can be either linear,quadratic or default');
end
end
% Normalize input
function set.assignments(self, assignments)
self.assignments = self.normalize_assignments(assignments);
end
function [energy,data_cost,pairwise_cost, data_term, pairwise_terms] = energy(self, assignments)
if (nargin < 2)
assignments = self.assignments;
else
assignments = self.normalize_assignments(assignments);
end
if (nargout < 4)
[energy, data_cost, pairwise_cost] = ...
energy_mex(...
self.cost_function, ...
self.dimensions, ...
assignments, ...
self.points, ...
self.connectivity - 1, ...
self.neighborhood.value, ...
self.data_weight, ...
self.tol, ...
self.eps, ...
self.verbose);
else
[energy, data_cost, pairwise_cost, data_term, pairwise_terms] = ...
energy_mex(...
self.cost_function, ...
self.dimensions, ...
assignments, ...
self.points, ...
self.connectivity - 1, ...
self.neighborhood.value, ...
self.data_weight, ...
self.tol, ...
self.eps, ...
self.verbose);
end
end
function num = num_pairwise(self)
num = numel(self.neighborhood.ind1);
end
function local_optimization(self, max_iterations)
if nargin < 2
max_iterations = 5000;
end
new_assignments = local_optimization_mex( ...
self.cost_function, ...
self.dimensions, ...
self.assignments, ...
self.points, ...
self.connectivity - 1, ...
self.neighborhood.value, ...
self.data_weight, ...
self.tol, ...
self.eps, ...
self.verbose, ...
int32(max_iterations));
if (self.energy(new_assignments) < self.energy)
self.assignments = new_assignments;
end
end
function conn = connectivity(self)
conn = uint32([self.neighborhood.ind1'; self.neighborhood.ind2']);
end
% Some local optimization and fusion
function optimize(self, iterations, num_global_props)
if nargin < 3
num_global_props = 500;
end
for iter = 1:iterations
% Generate some proposals
mean_local = self.generate_mean_proposals();
mean_random = self.generate_global_proposals(num_global_props);
%% Optimize
self.local_optimization;
self.batch_binary_fusion(mean_local);
self.batch_binary_fusion(mean_random);
if (self.verbose)
self.plot();
drawnow();
end
end
end
% Geneate new proposals by averaging over neighbors
function proposals = generate_mean_proposals(self)
% Intiallize with current assignments
% for neighborhood with less than max_neighor neighbors
% this will be the placeholder assigment.
proposals = cell(self.max_neighbors,1);
for i = 1:self.max_neighbors
proposals{i} = self.assignments;
end
for i = 1:self.num_points
%Mean of neghboring planes
p = 0;
for j = self.neighborhood.ind2(self.neighborhood.ind1 == i)';
p = p+1;
p1 = self.points(:,i);
p2 = self.points(:,j);
q = (p1+p2)./2;
n1 = self.assignments(1:end-1,i);
n2 = self.assignments(1:end-1,j);
d2 = self.assignments(end,j);
if abs(acos(n1'*n2)) > pi/2
n2 = -self.assignments(1:end-1,j);
d2 = -self.assignments(end,j);
end
dm2 = -(d2+q'*n2);
nnew = mean([n1 n2],2);
nnew = nnew./norm(nnew);
dnew = -dm2-nnew'*q;
proposals{p}(:,i) = [nnew; dnew];
end
end
end
end
methods (Access = protected)
function assignments = normalize_assignments(~, assignments)
denom = sqrt(sum(assignments(1:end-1,:).^2,1));
assignments = assignments./repmat(denom,[size(assignments,1) 1]);
end
end
end