/
modd2_visualize_frame.m
259 lines (199 loc) · 13.1 KB
/
modd2_visualize_frame.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
function [vis_img, tp, fp, fn] = modd2_visualize_frame(dataset_path, segmentation_output_path, method_name, author_name, sequence_number, frame_number, rectified, for_video_bool, segmentation_colors, tp_in, fp_in, fn_in)
if(nargin < 7)
error('Not enough input parameters');
elseif(nargin < 8)
for_video_bool = 0;
tp_in = 0;
fp_in = 0;
fn_in = 0;
segmentation_colors = [0, 255, 0; ...
0, 0, 0; ...
255, 0, 0];
elseif(nargin < 9)
segmentation_colors = [0, 255, 0; ...
0, 0, 0; ...
255, 0, 0];
tp_in = 0;
fp_in = 0;
fn_in = 0;
end
%% Get evaluation parameters
eval_params = get_eval_params(segmentation_colors, rectified);
% Add sequence ID
seq.id = sequence_number;
seq.mask_num = frame_number;
% Get sequence details (name, start frame, end frame)
[seq.name, seq.start_frame, seq.end_frame] = get_seq_details(sequence_number);
%% Get paths
paths = get_paths(dataset_path, segmentation_output_path, method_name, seq, eval_params);
[vis_img, tp, fp, fn] = perform_visualization(paths, seq, eval_params, method_name, author_name, for_video_bool, tp_in, fp_in, fn_in);
end
function [vis_img, tp, fp, fn] = perform_visualization(paths, seq, eval_params, method_name, author_name, for_video_bool, tp_in, fp_in, fn_in)
% false from boat detections - discard this....
%fp_mask = imread('rectf_mask_fp.png');
%fp_mask = imdilate(imbinarize(imresize(fp_mask, [384, 512], 'nearest')), strel('disk', 4, 4));
% Load chosen raw image from left camera
raw_img = imread(fullfile(paths.frames, sprintf('%08dL.jpg', seq.start_frame + seq.mask_num)));
% Load segmentation output for the chosen frame
cnn_seg_msk = imread(fullfile(paths.segmentation_output, sprintf('%08dL_pred.png', seq.start_frame + seq.mask_num)));
% Mask parts of the USV, rectify the segmentation mask if needed and
% mask the black/nan areas caused by the rectification...
cnn_seg_msk = mask_parts_of_usv(cnn_seg_msk, paths, seq, eval_params);
% Postprocess output segmentation images and output extracted sea
% mask and detected objects...
[det_objs, sea_mask] = postprocess_output_image(cnn_seg_msk, eval_params);
% Prettyfy segmentation mask
cnn_seg_msk = change_colors_img(cnn_seg_msk, eval_params);
% Overlay segmentation mask over image
segmented_image = image_overlay_segmentation_mask(raw_img, cnn_seg_msk, sea_mask);
% QR code to modd2 project page (do not remove!)
qr_code = imread('viz_images/qr-code_2_bw.png');
qr_size_y = 180;
qr_size_x = 180;
qr_code = imresize(qr_code, [qr_size_y, qr_size_x]);
qr_code = imbinarize(qr_code);
% Load ground truth file
gtl = load(fullfile(paths.ground_truth, sprintf('%08dL.mat', seq.start_frame + seq.mask_num)));
gtl = gtl.annotations;
%filter gt sea edge
%sea_edge_line = gtl.sea_edge( logical((gtl.sea_edge(:,1) > 0) .* (gtl.sea_edge(:,1) <= eval_params.img_size(2))), : );
% Ignore coordinates with inf or nan values...
sea_edge_line = filter_sea_edge(gtl.sea_edge, eval_params.img_size);
% create inverse sea mask
tmp_inv_sea_mask = poly2mask([1; sea_edge_line(:,1); eval_params.img_size(2)], [1; sea_edge_line(:,2); 1], eval_params.img_size(1), eval_params.img_size(2));
% Filter obstacles
gtl = filter_obstacles(gtl, eval_params.img_size, tmp_inv_sea_mask);
% scaled raw image
scaled_image = imresize(raw_img, [320, 426]);
% scaled segmentation mask
scaled_seg_mask = imresize(cnn_seg_msk, [320, 426]);
% Create image
final_img = zeros(eval_params.img_size(1), eval_params.img_size(2), 3);
offset_top = size(final_img, 1) - eval_params.img_size(1) + 1;
final_img(offset_top : end, :, :) = segmented_image;
if for_video_bool
% Raw input image subframe...
final_img(1:320, 1:426, :) = scaled_image; % scaled raw image top left
final_img = insertText(final_img, [213 1], 'Input image', 'FontSize', 25, 'TextColor', [255, 255, 255], 'BoxOpacity', 0.15, 'AnchorPoint', 'CenterTop');
% Postprocessed segmentation mask subframe...
final_img(1:320, 427:852, :) = scaled_seg_mask; % scaled raw segmentation mask top middle
final_img = insertText(final_img, [639 1], 'Segmentation mask', 'FontSize', 25, 'TextColor', [255, 255, 255], 'BoxOpacity', 0.15, 'AnchorPoint', 'CenterTop');
% QR code..
for i = 1 : 3
final_img(1:320, 853:end, i) = final_img(1:320, 853:end, i) * 0.7 + 0.3 * zeros(320, 426, 1);
tmp_qr_background = final_img(320 - qr_size_y + 1 : 320, size(final_img, 2) - qr_size_x + 1 : end, i);
tmp_qr_background(qr_code > 0) = qr_code(qr_code > 0) * 255;
final_img(320 - qr_size_y + 1 : 320, size(final_img, 2) - qr_size_x + 1 : end, i) = tmp_qr_background;
end
final_img(321, :, :) = 0;
final_img = insertText(final_img, [size(final_img, 2) - qr_size_x + 1, 320 - qr_size_y + 10], 'MODD2 dataset:', 'FontSize', 19, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftBottom');
% Authors and Method name
final_img = insertText(final_img, [853, 1], sprintf('Method: %*.*s', 0, 30, method_name), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
final_img = insertText(final_img, [853, 40], sprintf('Authors: %*.*s', 0, 29, author_name), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
end
% Footer method name
final_img = insertText(final_img, [1, 938], method_name, 'FontSize', 12, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
% Evaluate
% Segmentation metrics
paths.gt_water_masks = 'ground_truth_water_map';
paths.dt_water_masks = fullfile('results', method_name, 'postprocessing');
% Obstacle detection metrics
[rmse_water, tp, fp, fn, tp_list, fp_list, fn_list] = evaluate_detections_modd2(sea_mask, det_objs, gtl, eval_params);
if for_video_bool
final_img = insertText(final_img, [853, 140], sprintf('W. Edge: %02.01fpx [%.01f%%]', (rmse_water * size(final_img, 2)), (rmse_water * 100)), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
final_img = insertText(final_img, [853, 175], sprintf('Total TP: %d', tp + tp_in), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
final_img = insertText(final_img, [853, 210], sprintf('Total FP: %d', fp + fp_in), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
final_img = insertText(final_img, [853, 245], sprintf('Total FN: %d', fn + fn_in), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
final_img = insertText(final_img, [853, 280], sprintf('Total F1: %.01f%%', ( (2*(tp+tp_in)) / (2*(tp+tp_in)+(fp+fp_in)+(fn+fn_in)) ) * 100), 'FontSize', 22, 'TextColor', [255, 255, 255], 'BoxOpacity', 0, 'AnchorPoint', 'LeftTop');
end
final_img = uint8(final_img);
figure(1);
clf; imagesc(final_img); axis equal; axis tight; hold on;
%plot(sea_edge_line(:,1), offset_top + sea_edge_line(:,2), 'LineWidth', 3, 'color', [0,0,0]); hold on;
%plot(sea_edge_line(:,1), offset_top + sea_edge_line(:,2), 'LineWidth', 2, 'color', [1,0,1]); hold on;
draw_all_obstacles(gtl, tp_list, fp_list, fn_list, eval_params.area_threshold);
set(gca,'xtick',[]);
set(gca,'ytick',[]);
set(gca,'xticklabel',[]);
set(gca,'yticklabel',[]);
%set(gca,'LooseInset',get(gca,'TightInset'));
%set(gca, 'FontSmoothing', 'on')
%set(gcf,'Position',[0 0 1278 958])
set(gca,'position',[0 0 1 1],'units','normalized')
vis_img = getframe(gca);%final_img;
end
function new_im = change_colors_img(old_im, eval_params)
[size_y, size_x, size_c] = size(old_im);
new_im = zeros(size_y, size_x, size_c);
%sky
new_im(:,:,1) = new_im(:,:,1) + (old_im(:,:,1) == eval_params.labels(1,1) & old_im(:,:,2) == eval_params.labels(1,2) & old_im(:,:,3) == eval_params.labels(1,3)) * 69;
new_im(:,:,2) = new_im(:,:,2) + (old_im(:,:,1) == eval_params.labels(1,1) & old_im(:,:,2) == eval_params.labels(1,2) & old_im(:,:,3) == eval_params.labels(1,3)) * 78;
new_im(:,:,3) = new_im(:,:,3) + (old_im(:,:,1) == eval_params.labels(1,1) & old_im(:,:,2) == eval_params.labels(1,2) & old_im(:,:,3) == eval_params.labels(1,3)) * 161;
%ground
new_im(:,:,1) = new_im(:,:,1) + (old_im(:,:,1) == eval_params.labels(2,1) & old_im(:,:,2) == eval_params.labels(2,2) & old_im(:,:,3) == eval_params.labels(2,3)) * 246;
new_im(:,:,2) = new_im(:,:,2) + (old_im(:,:,1) == eval_params.labels(2,1) & old_im(:,:,2) == eval_params.labels(2,2) & old_im(:,:,3) == eval_params.labels(2,3)) * 193;
new_im(:,:,3) = new_im(:,:,3) + (old_im(:,:,1) == eval_params.labels(2,1) & old_im(:,:,2) == eval_params.labels(2,2) & old_im(:,:,3) == eval_params.labels(2,3)) * 59;
%water
new_im(:,:,1) = new_im(:,:,1) + (old_im(:,:,1) == eval_params.labels(3,1) & old_im(:,:,2) == eval_params.labels(3,2) & old_im(:,:,3) == eval_params.labels(3,3)) * 51;
new_im(:,:,2) = new_im(:,:,2) + (old_im(:,:,1) == eval_params.labels(3,1) & old_im(:,:,2) == eval_params.labels(3,2) & old_im(:,:,3) == eval_params.labels(3,3)) * 168;
new_im(:,:,3) = new_im(:,:,3) + (old_im(:,:,1) == eval_params.labels(3,1) & old_im(:,:,2) == eval_params.labels(3,2) & old_im(:,:,3) == eval_params.labels(3,3)) * 222;
new_im = uint8(new_im);
end
function im_out = image_overlay_segmentation_mask(img, processed_segm, sea_mask)
[size_y, size_x, ~] = size(img);
processed_segm_r = processed_segm(:,:,1);
processed_segm_g = processed_segm(:,:,2);
processed_segm_b = processed_segm(:,:,3);
sea_mask = logical(sea_mask);
processed_segm_r(sea_mask) = 0;
processed_segm_r = double(processed_segm_r) + sea_mask * 51;
processed_segm_g(sea_mask) = 0;
processed_segm_g = double(processed_segm_g) + sea_mask * 168;
processed_segm_b(sea_mask) = 0;
processed_segm_b = double(processed_segm_b) + sea_mask * 222;
processed_segm(:,:,1) = processed_segm_r;
processed_segm(:,:,2) = processed_segm_g;
processed_segm(:,:,3) = processed_segm_b;
processed_segm = double(imresize(uint8(processed_segm), [size_y, size_x], 'method', 'nearest'));
img_bw = rgb2gray(img);
im_out = zeros(size_y, size_x, 3);
w = [0.4, 0.4, 0.4];
for j = 1 : 3
tmp_img_bw = double(img_bw);
im_out(:,:,j) = tmp_img_bw * w(j) + (1 - w(j)) * processed_segm(:,:,j);
end
im_out = uint8(im_out);
end
function draw_all_obstacles(gt_obs, tp_list, fp_list, fn_list, area_threshold)
% Show large ground truth obstacles
plot(gt_obs.sea_edge(:,1), gt_obs.sea_edge(:,2), ':m', 'LineWidth', 2); hold on;
for i = 1 : size(gt_obs.largeobjects, 1)
tmp_rectangle = gt_obs.largeobjects(i, :);
tmp_rectangle(3) = tmp_rectangle(3) - tmp_rectangle(1);
tmp_rectangle(4) = tmp_rectangle(4) - tmp_rectangle(2);
rectangle('Position', tmp_rectangle, 'LineStyle', ':', 'LineWidth', 1, 'EdgeColor', 'k'); hold on;
end
% Show small ground truth obstacles
for i = 1 : size(gt_obs.smallobjects, 1)
tmp_rectangle = gt_obs.smallobjects(i, :);
tmp_rectangle(3) = tmp_rectangle(3) - tmp_rectangle(1);
tmp_rectangle(4) = tmp_rectangle(4) - tmp_rectangle(2);
% Plot small obstacle if it is large enough to be a threat
if(tmp_rectangle(4) * tmp_rectangle(3) <= area_threshold)
rectangle('Position', tmp_rectangle, 'LineStyle', ':', 'LineWidth', 1, 'EdgeColor', [1, 1, 1]); hold on;
end
end
% Show detected obstacles
for i = 1 : size(tp_list, 2)
rectangle('Position', [tp_list(1,i), tp_list(2,i), tp_list(3,i) - tp_list(1,i), tp_list(4,i) - tp_list(2,i)], 'LineWidth', 3, 'EdgeColor', 'k'); hold on;
rectangle('Position', [tp_list(1,i), tp_list(2,i), tp_list(3,i) - tp_list(1,i), tp_list(4,i) - tp_list(2,i)], 'LineWidth', 2, 'EdgeColor', 'g', 'FaceColor', [0, 1, 0, 0.25]); hold on;
end
for i = 1 : size(fp_list, 2)
rectangle('Position', [fp_list(1,i), fp_list(2,i), fp_list(3,i) - fp_list(1,i), fp_list(4,i) - fp_list(2,i)], 'LineWidth', 3, 'EdgeColor', 'k'); hold on;
rectangle('Position', [fp_list(1,i), fp_list(2,i), fp_list(3,i) - fp_list(1,i), fp_list(4,i) - fp_list(2,i)], 'LineWidth', 2, 'EdgeColor', [1, 0.65, 0]); hold on;
end
for i = 1 : size(fn_list, 2)
rectangle('Position', [fn_list(1,i), fn_list(2,i), fn_list(3,i) - fn_list(1,i), fn_list(4,i) - fn_list(2,i)], 'LineWidth', 3, 'EdgeColor', 'k'); hold on;
rectangle('Position', [fn_list(1,i), fn_list(2,i), fn_list(3,i) - fn_list(1,i), fn_list(4,i) - fn_list(2,i)], 'LineWidth', 2, 'EdgeColor', 'r', 'FaceColor', [1, 0, 0, 0.25]); hold on;
end
end