-
Notifications
You must be signed in to change notification settings - Fork 0
/
GenerateExampleData.py
243 lines (182 loc) · 9.34 KB
/
GenerateExampleData.py
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
import datetime
import sys
sys.path.insert(0, "/home/charlesrwest/cpp/projects/TensorFlow/objectTransform")
import bpy
import bmesh
from mathutils import Vector
from mathutils import Euler
from bpy_extras.object_utils import world_to_camera_view
import random
import math
import json
import time
import Parameters
import os
#cam = bpy.data.objects['Camera']
#obj = bpy.data.objects['Petshop-cat-figurine']
def RemoveImageFromMemory (passedName):
# Extra test because this can crash Blender.
img = bpy.data.images[passedName]
try:
img.user_clear()
can_continue = True
except:
can_continue = False
if can_continue == True:
try:
bpy.data.images.remove(img, True)
result = True
except:
result = False
else:
result = False
return result
def CollectBackgroundImagePaths(sourceDirectoryPath, imageExtensions):
file_paths = []
for root, dirs, files in os.walk(sourceDirectoryPath):
for file_name in files:
for extension in imageExtensions:
if file_name.endswith(extension):
file_paths.append(root + "/" + file_name)
return file_paths
def SetBackgroundImage(backgroundImagePath):
#Store reference to old image so it can be removed
old_image_name = bpy.data.materials[1].node_tree.nodes.get("Image Texture").image.name
#Load new image to be used
new_image = bpy.data.images.load(filepath = backgroundImagePath)
#Set the new image to be used
bpy.data.materials[1].node_tree.nodes.get("Image Texture").image = new_image
#Delete the old one from the file
remove_succeeded = RemoveImageFromMemory(old_image_name)
if not remove_succeeded:
print("Error removing " + old_image_name)
else:
print("Removing " + old_image_name + " worked")
#bpy.data.images[old_image_name].user_clear()
#bpy.data.images.remove(bpy.data.images[old_image_name])
def object_center_in_camera_view(camera, obj):
scene = bpy.context.scene
mat_world = obj.matrix_world
cs, ce = camera.data.clip_start, camera.data.clip_end
in_view = False
co_ndc = world_to_camera_view(scene, camera, obj.location)
in_view = (0.0 < co_ndc.x < 1.0 and 0.0 < co_ndc.y < 1.0 and cs < co_ndc.z < ce)
#print("Center view: " + str(in_view) + " with loc: " + str(obj.location) + " and camera rel position: " + str(Vector((co_ndc.x, co_ndc.y, co_ndc.z))))
return in_view
def location_in_camera_view(camera, location):
scene = bpy.context.scene
cs, ce = camera.data.clip_start, camera.data.clip_end
in_view = False
co_ndc = world_to_camera_view(scene, camera, location)
in_view = (0.0 < co_ndc.x < 1.0 and 0.0 < co_ndc.y < 1.0 and cs < co_ndc.z < ce)
#print("Center view: " + str(in_view) + " with loc: " + str(obj.location) + " and camera rel position: " + str(Vector((co_ndc.x, co_ndc.y, co_ndc.z))))
return in_view
def approx_object_in_camera_view(camera, centerLocation, marginFactor, distance, minDistance, maxDistance):
scene = bpy.context.scene
adjusted_margin = marginFactor*(maxDistance/distance)
cs, ce = camera.data.clip_start, camera.data.clip_end
in_view = False
co_ndc = world_to_camera_view(scene, camera, centerLocation)
in_view = (adjusted_margin < co_ndc.x < (1.0-adjusted_margin) and adjusted_margin < co_ndc.y < (1.0-adjusted_margin) and cs < co_ndc.z < ce)
#print("Center view: " + str(in_view) + " with loc: " + str(obj.location) + " and camera rel position: " + str(Vector((co_ndc.x, co_ndc.y, co_ndc.z))))
return in_view
def mesh_object_vertices_in_camera_view(camera, obj):
scene = bpy.context.scene
mat_world = obj.matrix_world
cs, ce = camera.data.clip_start, camera.data.clip_end
for vertex in obj.data.vertices:
co_ndc = world_to_camera_view(scene, camera, mat_world * vertex.co)
if(not (0.0 < co_ndc.x < 1.0 and 0.0 < co_ndc.y < 1.0 and cs < co_ndc.z < ce)):
#print("Vertex view: " + str(False))
return False
#print("Vertex view: " + str(True))
return True
def GetRandomUnitVector():
unit_vector = Vector((random.uniform(-1.0, 1.0), random.uniform(-1.0, 1.0), random.uniform(-1.0, 1.0)))
unit_vector.normalize()
return unit_vector
def PerturbInsideCameraView(minCenterDistance, maxCenterDistance, camera, obj):
relative_vector = Vector()
relative_euler_orientation = Euler()
relative_euler_orientation.x = random.uniform(-math.pi, math.pi)
relative_euler_orientation.y = random.uniform(-math.pi, math.pi)
relative_euler_orientation.z = random.uniform(-math.pi, math.pi)
#Find a position that is fully inside the camera's frustum
obj.rotation_mode = 'XYZ'
for i in range(0, 10000):
distance = random.uniform(minCenterDistance, maxCenterDistance)
relative_vector = GetRandomUnitVector()*distance
if not approx_object_in_camera_view(camera, camera.location + relative_vector, .15, distance, minCenterDistance, maxCenterDistance):
continue
bpy.context.scene.update()
#print("Updating scene " + str((time.time() - start )*1000.0) + " milliseconds")
#Temporarily disabling translation for debugging
obj.location = camera.location + relative_vector
obj.rotation_euler = relative_euler_orientation
#Update world matrix of object to match
bpy.context.scene.update()
#if not object_center_in_camera_view(camera, obj):
# continue
break
#start = time.time()
#if mesh_object_vertices_in_camera_view(camera, obj):
# print("Fine grain check " + str((time.time() - start )*1000.0) + " milliseconds")
# break
# print("Fine grain check " + str((time.time() - start )*1000.0) + " milliseconds")
return relative_vector, relative_euler_orientation
#Multiplies the location of object by the normalization factor before storing it in JSON
#Currently requires camera to be at origin without rotation
def GenerateExamples(numberOfExamples, objectName, cameraName, minDistance, maxDistance, locationNormalizationFactor, backgroundImagesDirectoryPath, outputDirectoryPath, includeLocation, includeRotation):
background_image_paths = CollectBackgroundImagePaths(backgroundImagesDirectoryPath, [".png", ".jpg"])
#Error on first render, so skip writing that one
bpy.ops.render.render( write_still=False )
results_map = dict()
for example_index in range(0, numberOfExamples):
current_frame_number = example_index+1
# if (example_index % 5) == 0:
# bpy.ops.wm.save_as_mainfile(filepath="/home/charlesrwest/Downloads/test_blend.blend")
if (example_index % 1000) == 0:
print("Generated " + str(example_index) + " images at " + datetime.datetime.now().strftime("%I:%M%p:%S on %B %d, %Y"))
if (example_index % 100000) == 0:
#Store labels in JSON file
# json_string = json.dumps(results_map, sort_keys=True, indent=4)
json_string = json.dumps(results_map)
json_file = open(outputDirectoryPath+"/labels.json", "w")
json_file.write(json_string)
json_file.close()
example_name = "example" + str(current_frame_number) + ".png"
while True:
try:
random_image_path = random.choice(background_image_paths)
SetBackgroundImage(random_image_path)
break
except:
pass
relative_vector, relative_euler_orientation = PerturbInsideCameraView(minDistance, maxDistance, bpy.data.objects[cameraName], bpy.data.objects[objectName])
relative_matrix_orientation = relative_euler_orientation.to_matrix()
#Render image
bpy.data.scenes['Scene'].render.filepath = outputDirectoryPath + "/" + example_name
bpy.ops.render.render( write_still=True )
#Store data for label (location, X axis, Y axis)
output_list = []
if(includeLocation):
output_list.append(relative_vector.x*locationNormalizationFactor)
output_list.append(relative_vector.y*locationNormalizationFactor)
output_list.append(relative_vector.z*locationNormalizationFactor)
if(includeRotation):
output_list.append(relative_matrix_orientation[0][0])
output_list.append(relative_matrix_orientation[1][0])
output_list.append(relative_matrix_orientation[2][0])
output_list.append(relative_matrix_orientation[0][1])
output_list.append(relative_matrix_orientation[1][1])
output_list.append(relative_matrix_orientation[2][1])
results_map[example_name] = output_list
print("Rendered " + example_name)
#Store labels in JSON file
#json_string = json.dumps(results_map, sort_keys=True, indent=4)
json_string = json.dumps(results_map)
json_file = open(outputDirectoryPath+"/labels.json", "w")
json_file.write(json_string)
json_file.close()
start = time.time()
GenerateExamples(Parameters.EXAMPLES_PER_EPOC, "Petshop-cat-figurine", "Camera", Parameters.MIN_OBJECT_CENTER_DISTANCE, Parameters.MAX_OBJECT_CENTER_DISTANCE, 1.0, "/home/charlesrwest/storage/Datasets/backgrounds/downloads", "/home/charlesrwest/storage/Datasets/objectTransform/rawData", Parameters.TRANSLATION_TRACKING_ENABLED, Parameters.ROTATION_TRACKING_ENABLED)