-
Notifications
You must be signed in to change notification settings - Fork 0
/
demo.m
195 lines (144 loc) · 8.68 KB
/
demo.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
% Demo function for presentation of our proposed method...
% Draws sea edge estimation and all detections on video
% along with some additional stuff...
% Input: Relative path to our dataset
function demo(dataset_path)
add_all_paths;
warning('off', 'all');
% demo video sequence...
% Note (you can change this to any number from interval [1,28]
%test_vid_number = 18;
test_vid_number = 7;
type_colorspace = Colorspace.rgb;
% Get video sequence details: - name of the sequence
% - name of the corresponding calibration sequence
% - start frame number
% - end frame number
[vid_seq_name, calib_seq_name, start_frame, end_frame, precalib_offset_one, precalib_offset_two] = getSequenceDetails(test_vid_number);
num_of_frames = end_frame - start_frame;
% Root directories paths
%images_root_dir = fullfile(dataset_path, 'video_data', vid_seq_name, 'frames');
images_root_dir = fullfile(dataset_path, 'video_data', vid_seq_name, 'framesRectified');
gt_root_dir = fullfile(dataset_path, 'annotations', vid_seq_name, 'groundTruth');
% Method functions: - imu_sensor (should the method use the IMU information?)
% - stereo_ver (should the method use stereo verification?)
% - scaled_image_size (segmentation image size in pixels)
% - num_cond_priors (which conditional priors to generate?)
imu_sensor = 1; % 0 = off, 1 = on
stereo_ver = 1; % 0 = off, 1 = on %1 orig
%scaled_image_size = [50, 50]; %orig
scaled_image_size = [100, 100];
num_cond_priors = 1; % 1 = sky and sea, 2 = sky only, 3 = sea only
% Load IMU sensor data if needed
if( imu_sensor )
% Load hiperpriors information
hiper_priors_fix = load('hiperprior-data/avgHorY.mat');
%load(fullfile(precalib_path, 'points.mat'));
%load(fullfile(precalib_path, 'imu1.mat'));
%load(fullfile(precalib_path, 'imu2.mat'));
fs = cv.FileStorage(fullfile(dataset_path, 'video_data', vid_seq_name, 'calibration.yaml'));
% Load imu calibration
% The IMU data format changes after sequence 15, because of the
% on-board software update...
if(test_vid_number < 15)
[B, ~, imu_initial_pitch, imu_initial_roll] = getGroundPlane( fullfile(dataset_path, 'video_data', calib_seq_name), 100);
else
[B, ~, imu_initial_roll, imu_initial_pitch] = getGroundPlane( fullfile(dataset_path, 'video_data', calib_seq_name), 100);
end
fprintf('IMU calibration loaded\n');
fprintf('IMU calibration files loaded...\n');
sim = [fs.imageSize{2}, fs.imageSize{1}];
end
% Get Fundamental matrix for stereo verification if needed
if( stereo_ver )
F = load(fullfile(dataset_path, 'video_data', calib_seq_name, 'fundamentalMatrix.mat')); %we load fundamental matrix from file...
end
% Get image size if "fs" file was not loaded
if(~exist('sim', 'var'))
sim = size( imread(fullfile( images_root_dir, sprintf( '%08dL.jpg', start_frame ))));
end
% Initialize detector state for left and right camera
detector_l.detector_state = DetectorState;
detector_r.detector_state = DetectorState;
fprintf('Processing video number %2d with options:\n\t*IMU sensor: %1d\n\t*Stereo verification: %1d\n\t*Image size: %d x %d\n', ...
test_vid_number, imu_sensor, stereo_ver, scaled_image_size(2), scaled_image_size(1));
% Timers initialization..
time_segm = 0;
time_stereo_ver = 0;
for fr = 1 : num_of_frames
% Load images
im_l = imread( fullfile( images_root_dir, sprintf( '%08dL.jpg', fr + start_frame ) ) );
im_r = imread( fullfile( images_root_dir, sprintf( '%08dR.jpg', fr + start_frame ) ) );
% Load ground truth files...
gt_l = load(fullfile(gt_root_dir, sprintf('%08dL.mat', fr + start_frame)));
%gt_l = gt_l.new_gt_l;
% load ground truth file for right camera in case of simple/mono
% segmentation
gt_r = load(fullfile(gt_root_dir, sprintf('%08dR.mat', fr + start_frame)));
%gt_r = gt_r.new_gt_r;
try
% Set colorspaces
detector_l.detector_state.colorspace = type_colorspace;
detector_r.detector_state.colorspace = type_colorspace;
% Build conditional priors if the IMU sensor is used
if( imu_sensor )
% Get horizon parameters...
[hor_line_param, ~] = getVanishingPoints(B, imu_initial_pitch, imu_initial_roll, fr + start_frame, fullfile(dataset_path, 'video_data', vid_seq_name), fs, test_vid_number, precalib_offset_one, precalib_offset_two);
% Get Horizon angle
hor_angle_l = atan(hor_line_param(1));
hor_angle_r = atan(hor_line_param(3));
% Get binary masks
[priors_sea_l, priors_sea_r, priors_sky_l, priors_sky_r] = imuMask(hor_line_param, scaled_image_size, sim, 3);
% Get y-position of horizon line's center
avg_hor_centery_l = ( round( sim(2) / 2 ) * hor_line_param(1) + hor_line_param(2) ) / sim(1);
avg_hor_centery_r = ( round( sim(2) / 2 ) * hor_line_param(3) + hor_line_param(4) ) / sim(1);
end
% Perform ISSM_s on the images...
detector_state_l = detector_l.detector_state;
detector_state_r = detector_r.detector_state;
tic;
%detector_l = water_perform_detection_on_image_ISSM(detector_l, im_l, 6.15, scaled_image_size, priors_sea_l, priors_sky_l, hor_angle_l, avg_hor_centery_l, hiper_priors_fix);
%detector_r = water_perform_detection_on_image_ISSM(detector_r, im_r, 6.15, scaled_image_size, priors_sea_r, priors_sky_r, hor_angle_r, avg_hor_centery_r, hiper_priors_fix);
[detector_state_l, sel_xy_l, objs_l, masked_sea_l, q_sum_large_l] = detect_edge_of_sea_simplified_ISSM(detector_state_l, im_l, 6.15, scaled_image_size, 'prior_sea', priors_sea_l, 'prior_sky', priors_sky_l, 'avg_hor_y', avg_hor_centery_l, 'hor_angle', hor_angle_l, 'hiper_fix', hiper_priors_fix);
[detector_state_r, sel_xy_r, objs_r, masked_sea_r, q_sum_large_r] = detect_edge_of_sea_simplified_ISSM(detector_state_r, im_r, 6.15, scaled_image_size, 'prior_sea', priors_sea_r, 'prior_sky', priors_sky_r, 'avg_hor_y', avg_hor_centery_r, 'hor_angle', hor_angle_r, 'hiper_fix', hiper_priors_fix);
time_segm_single = toc;
time_segm = time_segm + time_segm_single;
detector_l.detector_state = detector_state_l;
detector_r.detector_state = detector_state_r;
detector_l.objs = objs_l;
detector_r.objs = objs_r;
detector_l.sel_xy = sel_xy_l;
detector_r.sel_xy = sel_xy_r;
detector_l.maskedSea = masked_sea_l;
detector_r.maskedSea = masked_sea_r;
detector_l.is_initialized = true;
detector_r.is_initialized = true;
% Discard detections that are due to fisheye camera vignette or parts of the USV visible in camera's field of view
[tmp_objects_l, tmp_objects_r] = discardBoatDetections(dataset_path, detector_l.objs, detector_r.objs, sim, test_vid_number);
detector_l.objs = tmp_objects_l;
detector_r.objs = tmp_objects_r;
clear tmp_objects_l tmp_objects_r;
% Only used later on for visualization of obstacles discarded by the stereo verification method...
original_objs_l = detector_l.objs;
original_objs_r = detector_r.objs;
% Perform stere verification of the obstacles if enabled...
tic;
if( stereo_ver )
[tmp_objects_l, tmp_objects_r] = verifyObstacleConsistency(im_l, im_r, detector_l.objs, detector_r.objs, F);
detector_l.objs = tmp_objects_l;
detector_r.objs = tmp_objects_r;
clear tmp_objects_l tmp_objects_r;
end
time_stereo_ver_single = toc;
time_stereo_ver = time_stereo_ver + time_stereo_ver_single;
% Draw detection and water edge estimation...
drawDemo(im_l, q_sum_large_l, im_r, q_sum_large_r, priors_sea_l, priors_sea_r, priors_sky_l, priors_sky_r, hor_line_param, original_objs_l, original_objs_r, detector_l, detector_r, gt_l, gt_r, avg_hor_centery_l, avg_hor_centery_r, hiper_priors_fix, hor_angle_l, hor_angle_r); drawnow;
fprintf('Frame %03d done. Time for segmentation %0.3fs, time for stereo verification %0.3fs\n', fr, time_segm_single, time_stereo_ver_single);
catch err % Catch if any error occured (this shouldnt happend)
fprintf('Method crashed on frame %04d\n', fr);
errMsgTxt = getReport(err);
fprintf('%s\n', errMsgTxt);
end
end
fprintf('Demo finished....\n\tAverage time for segmentation: %1.3f\n\tAverage time for stereo verification: %1.3f\n\tFrame rate: %2.1f\n', (time_segm / num_of_frames), (time_stereo_ver / num_of_frames), 1 / ( (time_segm / num_of_frames) + (time_stereo_ver / num_of_frames) ) );
end