-
Notifications
You must be signed in to change notification settings - Fork 6
/
t_DatasetGenerate.m
419 lines (346 loc) · 14.9 KB
/
t_DatasetGenerate.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
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
%% Automatically assemble a country road scene
%
% Dependencies
% ISET3d, ISETAuto, ISETCam and scitran
% Prefix: ia- means isetauto
% pi- means iset3d-v4
%
% ISET3d-V4: Takes a PBRT file, parse 3D information including lights,
% materials, textures and meshes. Modify the properties and render it.
%
% ISETAuto: Assemble ISET3d OBJECT into a complex driving scene.
%
% ISETCam: Convert scene radiance or optical irradiance data to RGB
% image with a physically based sensor model and ISP pipeline.
%
% Zhenyi, 2022
%% Initialize ISET and Docker
ieInit;
if ~piDockerExists, piDockerConfig; end
camPositions = 2; % for each scene, we choose different cam positions.
sceneName = 'nightdrive0920';
fid_cpu = fopen(fullfile(piRootPath, 'local', sceneName, 'render_cpu.sh'),'w+');
fid_gpu_0 = fopen(fullfile(piRootPath, 'local', sceneName, 'render_gpu_0.sh'),'w+');
fid_gpu_1 = fopen(fullfile(piRootPath, 'local', sceneName, 'render_gpu_1.sh'),'w+');
fid_gpu_2 = fopen(fullfile(piRootPath, 'local', sceneName, 'render_gpu_2.sh'),'w+');
fid_gpu_3 = fopen(fullfile(piRootPath, 'local', sceneName, 'render_gpu_3.sh'),'w+');
%% Road initiation
% assetDir = fullfile(iaRootPath,'local','assets');
assetDir = iaFileDataRoot('type','PBRT_assets');
roadNames = {'road_001', 'road_002','road_003','road_004','road_005','road_006',...
'road_011','road_012','road_013','road_015','road_020'};
for ii = 1:25%numel(roadNames)
% roadName = roadNames{ii};
roadName = roadNames{randi(11)};
roadName = 'road_001';
roadDir = fullfile(iaFileDataRoot('type','PBRT_assets','road'),roadName);
% The road data
roadData = roadgen('road directory',roadDir, 'asset directory',assetDir);
assetLibNames = keys(assetlib());
%% Place the onroad elements
% The driving lanes
roadData.set('onroad car lanes',{'leftdriving','rightdriving'});
% roadData.set('onroad car lanes', {'leftdriving'});
% Cars on the road
carNames = assetLibNames(contains(assetLibNames,'car'));
roadData.set('onroad car names',carNames(randperm(numel(carNames), 8)));%
% How many cars on each driving lane.
% The vector length of these numbers should be the same as the number
% of driving lanes.
nCars = [randi(20), randi(20)];
roadData.set('onroad n cars', nCars);
truckNames = assetLibNames(contains(assetLibNames,'truck'));
roadData.onroad.truck.namelist = truckNames(randperm(numel(truckNames), 3));
roadData.onroad.truck.lane = {'leftdriving','rightdriving'};
roadData.onroad.truck.number = [2, 2];
busNames = assetLibNames(contains(assetLibNames,'bus'));
roadData.onroad.bus.namelist = busNames(randperm(numel(busNames), 3));
roadData.onroad.bus.lane = {'leftdriving','rightdriving'};
roadData.onroad.bus.number = [2, 2];
bikerNames = assetLibNames(contains(assetLibNames,'biker'));
roadData.onroad.biker.namelist = bikerNames(randperm(numel(bikerNames), 7));
roadData.onroad.biker.lane = {'leftdriving','rightdriving'};
roadData.onroad.biker.number = [3, 3];
pedestrianNames = assetLibNames(contains(assetLibNames,'pedestrian'));
roadData.onroad.pedestrian.namelist = pedestrianNames(randperm(numel(pedestrianNames), 30));
roadData.onroad.pedestrian.lane = {'leftdriving','rightdriving'};
roadData.onroad.pedestrian.number = [5, 5];
% Now place the animals
deerNames = assetLibNames(contains(assetLibNames,'deer'));
deerNames = deerNames(randperm(numel(deerNames),2));
roadData.set('onroad animal names',deerNames);
roadData.set('onroad n animals', [randi(3)-1, randi(3)-1]);
roadData.set('onroad animal lane',{'leftdriving','rightdriving'});
% roadData.onroad.animal.namelist = {'deer_001'};
% roadData.onroad.animal.number= randi(10);
% roadData.onroad.animal.lane = {'rightdriving'};
%% Place the offroad elements. These are animals and trees. Not cars.
roadData.set('offroad animal names',deerNames);
roadData.set('offroad n animals', [randi(6)-1, randi(6)-1]);
roadData.set('offroadanimallane', {'rightshoulder','leftshoulder'});
roadData.set('offroad animal min distance',0);
roadData.set('offroad animal layer width',5);
pedestrianNames = assetLibNames(contains(assetLibNames,'pedestrian'));
roadData.offroad.pedestrian.namelist = pedestrianNames(randperm(numel(pedestrianNames), 30));
roadData.offroad.pedestrian.lane = {'rightshoulder','leftshoulder'};
roadData.offroad.pedestrian.number = [10, 10];
treeNames = assetLibNames(contains(assetLibNames,'tree'));
treeNames = treeNames(randperm(numel(treeNames), 10));
roadData.set('offroad tree names', treeNames);
roadData.set('offroad n trees', [100, 200, 30]); % [50, 100, 150]
roadData.set('offroad tree lane', {'rightshoulder','leftshoulder'});
% Place the trees for different distance range from the boundary of road
roadData.offroad.grass.namelist = {'grass_001','grass_002','grass_003','grass_004','grass_005','grass_006','grass_007','grass_008'};
roadData.offroad.grass.number = [1000, 50, 10];%[800, 100, 50];
roadData.offroad.grass.lane = {'rightshoulder','leftshoulder'};
roadData.offroad.streetlight.namelist = {'streetlight_001'};
roadData.offroad.streetlight.number = [10, 10];
roadData.offroad.streetlight.lane = {'rightshoulder', 'leftshoulder'};
roadData.offroad.rock.namelist = {'rock_001','rock_002','rock_003'};
roadData.offroad.rock.number = [200, 200, 1];
roadData.offroad.rock.lane = {'rightshoulder','leftshoulder'};
%% Set the recipe parameters
thisR = roadData.recipe;
thisR.set('film render type',{'radiance','depth'});
% render quality
thisR.set('film resolution',[1920 1080]); % 4
thisR.set('pixel samples',128); % 512
thisR.set('max depth',5);
thisR.set('sampler subtype','pmj02bn');
imageID = iaImageID();
% outputFile = fullfile(iaRootPath, 'local', sceneName, [num2str(imageID),'.pbrt']);
outputFile = fullfile(piRootPath, 'local', sceneName, [num2str(imageID),'.pbrt']);
thisR.set('outputFile',outputFile);
%% Set up the rendering skymap
skymapLists = dir([fullfile(iaFileDataRoot('type','PBRT_assets'), 'skymap'), ...
filesep(), 'sky*.exr']);
skymapRandIndex = randi(size(skymapLists,1));
skymapName = skymapLists(skymapRandIndex).name;
% skymapName = 'sky-noon_009.exr';
thisR.set('skymap',fullfile(skymapLists(skymapRandIndex).folder, skymapName));
thisR.set('asset',strrep(skymapName,'.exr','_B'),'rotation', [0 0 0]);
% useful Docker cmd for reading or making a skymap.
%{
piDockerImgtool('makeequiarea','infile','/Users/zhenyi/git_repo/dev/iset3d-v4/data/lights/dikhololo_night_4k.exr');
%}
%% Assemble the scene using ISET3d methods
assemble_tic = tic();
roadData.assemble();
fprintf('---> Scene assembled in %.f seconds.\n',toc(assemble_tic));
%% Apply our customized material
iaAutoMaterialGroupAssignV4(thisR, true);
disp('--> Material assigned');
%% Use a camera for this car
% lensfile = 'wide.40deg.6.0mm.json'; % 30 38 18 10
% fprintf('Using lens: %s\n',lensfile);
% random pick a car, use the camera on it. This are the types of cameras
% so far:
%
% front
% back
% left
% camera_type = 'right'
for cc = 1:camPositions
camera_type = 'front';
% random pick a car, use the camera on it.
branchID(cc) = roadData.cameraSet(camera_type); % (camera_type, car_id)
if cc > 1 && branchID(cc) == branchID(cc-1)
disp('Cam position was used, pick a different.');
branchID(cc) = roadData.cameraSet('camera type', camera_type); % (camera_type, car_id)
end
direction = thisR.get('object direction');
thisR.set('object distance', 0.95);
%% Render the scene, and maybe an OI
if cc > 1
imageID = iaImageID();
outputFile = fullfile(piRootPath, 'local', sceneName, [num2str(imageID),'.pbrt']);
thisR.set('outputFile',outputFile);
end
outputFile = thisR.get('output file');
sceneRecipe = strrep(outputFile,'.pbrt','.mat');
save(sceneRecipe,'thisR','-mat');
%% create light group
skyName = erase(skymapName,'.exr');
% recipeList = iaLightsGroup(thisR, skyName);
recipeList{1} = thisR;
for rr = 1:numel(recipeList)
piWrite(recipeList{rr});
gpuDeviceID = randi(3);
[~, ~, renderCMD] = piRenderZhenyi(recipeList{rr}, 'gpuDeviceID', gpuDeviceID, ...
'renderLater',true);
switch gpuDeviceID
case 0
fprintf(fid_gpu_0, [renderCMD,'\n']);
case 1
fprintf(fid_gpu_1, [renderCMD,'\n']);
case 2
fprintf(fid_gpu_2, [renderCMD,'\n']);
case 3
fprintf(fid_gpu_3, [renderCMD,'\n']);
end
end
%% render label
% Label the objects using the CPU
% renderLater = true;
% [objectslist, instanceIdMap, renderCMD] = piRenderLabel(thisR, renderLater); % (thisR, renderLater)
% fprintf(fid_cpu, [renderCMD,'\n']);
end
end
% run command in background
fclose(fid_gpu_0);
fclose(fid_gpu_1);
fclose(fid_gpu_2);
fclose(fid_gpu_3);
fclose(fid_cpu);
%{
for campos = 1:5
camera_type = 'front';
% random pick a car, use the camera on it.
branchID = roadData.cameraSet(camera_type); % (camera_type, car_id)
direction = thisR.get('object direction');
thisR.set('object distance', 0.95);
imageID = iaImageID();
outputFile = fullfile(piRootPath, 'local', sceneName, [num2str(imageID),'.pbrt']);
thisR.set('outputFile',outputFile);
piWrite(thisR);
scene = piRenderZhenyi(thisR);sceneWindow(scene);
outputFile = thisR.get('output file');
sceneRecipe = strrep(outputFile,'.pbrt','.mat');
save(sceneRecipe,'thisR','-mat');
end
%% create light group
skyName = erase(skymapName,'.exr');
recipeList = iaLightsGroup(thisR, skyName);
%%
for rr = 1:numel(recipeList)
% recipeList{rr}.set('pixel samples',1024);
% recipeList{rr}.set('film resolution',[1280 720]*1.5);
piWrite(recipeList{rr});
scene_lg{rr} = piRenderZhenyi(recipeList{rr}, 'meanluminance',0);
end
%{
%debug: scene = piRenderZhenyi(thisR);sceneWindow(scene);
% The position does not seem to change correctly yet.
% We do have a repeatable scene if we change from front - left -
% front, we get the same scene back. But the 'left' position doesn't
% seem good.
camera_type = 'front';
roadData.cameraSet(camera_type, branchID); % (camera_type, car_id)
[scene, res] = piWRS(thisR);
%}
%{
oi = oiCreate;
oi = oiCompute(oi,scene);
oi = oiCrop(oi,'border');
sensor = sensorCreate('MT9V024');
sensor = sensorSet(sensor,'fov',sceneGet(scene,'fov'),oi);
sensor = sensorSet(sensor,'auto exposure',true);
% sensor = sensorSet(sensor,'exposure time',0.016);
sensor = sensorCompute(sensor,oi);
ip = ipCreate;
ip = ipCompute(ip, sensor);
% ipWindow(ip);
%}
for rr = 1:3
scene_lg_dn{rr} = piAIdenoise(scene_lg{rr});
end
%% Label the objects using the CPU
[objectslist, instanceIdMap] = piRenderLabel(thisR);
% [objectslist,instanceMap] = roadData.label();
%% Show the various images
ieNewGraphWin([],'upperleftbig');
% We should be able to use the sensor image for finding the objects.
% But not yet.
imgscene = sceneGet(scene,'rgb');
% imgscene = ipGet(ip,'srgb');
subplot(2,2,1);
imshow(imgscene);title('Radiance')
ax1 = subplot(2,2,2);
imagesc(scene.depthMap);colormap(ax1,"gray");title('Depth');axis off
set(gca, 'Visible', 'off');
ax2=subplot(2,2,3);
imagesc(instanceMap);colormap(ax2,"colorcube");axis off;title('Pixel Label');
subplot(2,2,4);
imshow(imgscene);title('Bounding Box');
%% Add the bounding boxes, which requires the cocoapi method
nBox=1;
nImage = 1;
Annotation=[];
[h,w,~] = size(imgscene);
datasetFolder = fullfile(piRootPath,'local','dataset_demo');
% write out object ID for segmentation map;
if ~exist(fullfile(datasetFolder,'additionalInfo'),'dir')
mkdir(fullfile(datasetFolder,'additionalInfo'))
end
seg_FID = fopen(fullfile(datasetFolder,'additionalInfo',[num2str(imageID),'.txt']),'w+');
fprintf(seg_FID,'sceneName: %s\nSkymap: %s\nCameraType: %s\n',sceneName, ...
erase(skymapName,'.exr'), camera_type);
fprintf(seg_FID,'Object ID:\n');
for ii = 1:numel(objectslist)
name = objectslist{ii};
name = erase(name,{'ObjectInstance ', '"', '_m'});
fprintf(seg_FID, '%d %s \n',ii, name);
if contains(lower(name), {'car'})
label = 'vehicle';
catId = 3;
r = 0.1; g= 0.5; b = 0.1;
elseif contains(lower(name),'deer')
label = 'Deer';
catId = 9;
r = 1; g= 0.1; b = 0.1;
else
continue;
end
[occluded, truncated, bbox2d, segmentation, area] = piAnnotationGet(instanceMap,ii,0);
if isempty(bbox2d), continue;end
pos = [bbox2d.xmin bbox2d.ymin ...
bbox2d.xmax-bbox2d.xmin ...
bbox2d.ymax-bbox2d.ymin];
rectangle('Position',pos,'EdgeColor',[r g b],'LineWidth',1);
tex=text(bbox2d.xmin+2.5,bbox2d.ymin-8,label);
tex.Color = [1 1 1];
tex.BackgroundColor = [r g b];
tex.FontSize = 12;
Annotation_coco{nBox} = struct('segmentation',segmentation,'area',area,'iscrowd',0,...
'image_id',sprintf('%d',imageID),'bbox',pos,'category_id',catId,'id',0,'ignore',0); %#ok<SAGROW>
fprintf('Class %s, instanceID: %d \n', label, ii);
nBox = nBox+1;
end
truesize;
%% Save the images
%{
% We are going to put the rgb image, depth map, pixel label, and
% bounding box in COCO format using this directory. You can use these
% image data again later.
datasetFolder = fullfile(iaRootPath,'local','nightdrive','dataset');
if ~exist(datasetFolder,'dir'), mkdir(datasetFolder); end
if ~exist(fullfile(datasetFolder,'rgb'),'dir')
mkdir(fullfile(datasetFolder,'rgb'))
end
if ~exist(fullfile(datasetFolder,'segmentation'),'dir')
mkdir(fullfile(datasetFolder,'segmentation'))
end
if ~exist(fullfile(datasetFolder,'depth'),'dir')
mkdir(fullfile(datasetFolder,'depth'))
end
if ~exist(fullfile(datasetFolder,'rendered'),'dir')
mkdir(fullfile(datasetFolder,'rendered'))
end
imgName = sprintf('%d.png',imageID);
% Image_coco = struct('file_name',imgName,'height',h,'width',w,'id',sprintf('%d',imageID));
%
% % write files out
% save(fullfile(datasetFolder, sprintf('%d_image.mat',imageID)),'Image_coco');
% save(fullfile(datasetFolder, sprintf('%d_anno.mat',imageID)), 'Annotation_coco');
imgFilePath = fullfile(datasetFolder,'rgb',imgName);
imwrite(radiance,imgFilePath);
imwrite(uint16(instanceMap),fullfile(datasetFolder,'segmentation',imgName));
imwrite(uint16(scene.depthMap),fullfile(datasetFolder,'depth',imgName));
outputFolder = roadData.recipe.get('outputdir');
movefile(fullfile(outputFolder,sprintf('renderings/%d.exr',imageID)),fullfile(datasetFolder,'rendered/'));
%}
% fprintf('****** Scene%d Generated! ******\n',nScene);
% end
%}
%% End
%}