Hugging Face's logo

qicq1c
/
SpatialCode

Model card Files
xet
 Community
qicq1c commited on
Commit
0ccb946
·
verified ·
1 Parent(s): 000efa3

Add scene0251_00 quick start example

Browse files
Files changed (4) hide show
  1. examples/scene0251_00/README.md +7 -0
  2. examples/scene0251_00/qa.json +90 -0
  3. examples/scene0251_00/spatial_code.json +200 -0
  4. examples/scene0251_00/vqa_input.json +1483 -0
examples/scene0251_00/README.md ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ # scene0251_00 Quick Start Example
2
+
3
+ This folder contains a lightweight SpatialCode example for ScanNet scene0251_00.
4
+
5
+ - `spatial_code.json`: predicted scene-level spatial code.
6
+ - `qa.json`: VQA questions for this scene.
7
+ - `vqa_input.json`: questions merged with spatial code, ready for `Train/RL/scripts/run_code_to_answer.py`.
examples/scene0251_00/qa.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "id": 3761,
4
+ "dataset": "scannet",
5
+ "scene_name": "scene0251_00",
6
+ "question_type": "room_size_estimation",
7
+ "question": "What is the size of this room (in square meters)? \nIf multiple rooms are shown, estimate the size of the combined space.",
8
+ "ground_truth": "35.9",
9
+ "options": null,
10
+ "key_object": []
11
+ },
12
+ {
13
+ "id": 4386,
14
+ "dataset": "scannet",
15
+ "scene_name": "scene0251_00",
16
+ "question_type": "object_counting",
17
+ "question": "How many chair(s) are in this room?",
18
+ "ground_truth": "2",
19
+ "options": null,
20
+ "key_object": [
21
+ "chair"
22
+ ]
23
+ },
24
+ {
25
+ "id": 4387,
26
+ "dataset": "scannet",
27
+ "scene_name": "scene0251_00",
28
+ "question_type": "object_counting",
29
+ "question": "How many window(s) are in this room?",
30
+ "ground_truth": "3",
31
+ "options": null,
32
+ "key_object": [
33
+ "window"
34
+ ]
35
+ },
36
+ {
37
+ "id": 4388,
38
+ "dataset": "scannet",
39
+ "scene_name": "scene0251_00",
40
+ "question_type": "object_counting",
41
+ "question": "How many door(s) are in this room?",
42
+ "ground_truth": "3",
43
+ "options": null,
44
+ "key_object": [
45
+ "door"
46
+ ]
47
+ },
48
+ {
49
+ "id": 4389,
50
+ "dataset": "scannet",
51
+ "scene_name": "scene0251_00",
52
+ "question_type": "object_counting",
53
+ "question": "How many trash bin(s) are in this room?",
54
+ "ground_truth": "5",
55
+ "options": null,
56
+ "key_object": [
57
+ "trash bin"
58
+ ]
59
+ },
60
+ {
61
+ "id": 4390,
62
+ "dataset": "scannet",
63
+ "scene_name": "scene0251_00",
64
+ "question_type": "object_counting",
65
+ "question": "How many table(s) are in this room?",
66
+ "ground_truth": "2",
67
+ "options": null,
68
+ "key_object": [
69
+ "table"
70
+ ]
71
+ },
72
+ {
73
+ "id": 4994,
74
+ "dataset": "scannet",
75
+ "scene_name": "scene0251_00",
76
+ "question_type": "route_planning",
77
+ "question": "You are a robot beginning at the poster and facing the cabinet. You want to navigate to the trashbin. You will perform the following actions (Note: for each [please fill in], choose either 'turn back,' 'turn left,' or 'turn right.'): 1. Go forward until the cabinet. 2. [please fill in] 3. Go forward until the trashbin. You have reached the final destination.",
78
+ "ground_truth": "A",
79
+ "options": [
80
+ "A. Turn Left",
81
+ "B. Turn Right",
82
+ "C. Turn Back"
83
+ ],
84
+ "key_object": [
85
+ "poster",
86
+ "cabinet",
87
+ "trashbin"
88
+ ]
89
+ }
90
+ ]
examples/scene0251_00/spatial_code.json ADDED
@@ -0,0 +1,200 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "scene0251_00": [
3
+ {
4
+ "bbox_3d": [
5
+ 4.428,
6
+ 7.408,
7
+ 1.642,
8
+ 0.455,
9
+ 0.258,
10
+ 0.931,
11
+ 0.0,
12
+ 0.0,
13
+ -1.395
14
+ ],
15
+ "label": "window"
16
+ },
17
+ {
18
+ "bbox_3d": [
19
+ 4.209,
20
+ 6.806,
21
+ 0.962,
22
+ 1.357,
23
+ 0.736,
24
+ 1.514,
25
+ 0.0,
26
+ 0.0,
27
+ -1.492
28
+ ],
29
+ "label": "table"
30
+ },
31
+ {
32
+ "bbox_3d": [
33
+ 3.927,
34
+ 7.236,
35
+ 1.853,
36
+ 0.165,
37
+ 0.575,
38
+ 1.046,
39
+ 0.0,
40
+ 0.0,
41
+ -1.37
42
+ ],
43
+ "label": "window"
44
+ },
45
+ {
46
+ "bbox_3d": [
47
+ 1.194,
48
+ 4.451,
49
+ 1.528,
50
+ 0.459,
51
+ 0.533,
52
+ 0.447,
53
+ 0.0,
54
+ 0.0,
55
+ -1.655
56
+ ],
57
+ "label": "trash bin"
58
+ },
59
+ {
60
+ "bbox_3d": [
61
+ 0.993,
62
+ 3.58,
63
+ 1.528,
64
+ 0.497,
65
+ 0.657,
66
+ 0.434,
67
+ 0.0,
68
+ 0.0,
69
+ -1.484
70
+ ],
71
+ "label": "trash bin"
72
+ },
73
+ {
74
+ "bbox_3d": [
75
+ 0.626,
76
+ 3.272,
77
+ 1.709,
78
+ 0.412,
79
+ 0.619,
80
+ 0.421,
81
+ 0.0,
82
+ 0.0,
83
+ -1.548
84
+ ],
85
+ "label": "trash bin"
86
+ },
87
+ {
88
+ "bbox_3d": [
89
+ 0.654,
90
+ 2.783,
91
+ 1.572,
92
+ 0.519,
93
+ 0.664,
94
+ 0.526,
95
+ 0.0,
96
+ 0.0,
97
+ -1.618
98
+ ],
99
+ "label": "trash bin"
100
+ },
101
+ {
102
+ "bbox_3d": [
103
+ 0.823,
104
+ 2.27,
105
+ 1.4,
106
+ 0.52,
107
+ 0.608,
108
+ 0.564,
109
+ 0.0,
110
+ 0.0,
111
+ -1.619
112
+ ],
113
+ "label": "trash bin"
114
+ },
115
+ {
116
+ "bbox_3d": [
117
+ 1.16,
118
+ 1.304,
119
+ 1.655,
120
+ 0.222,
121
+ 1.298,
122
+ 0.781,
123
+ 0.0,
124
+ 0.0,
125
+ -1.693
126
+ ],
127
+ "label": "door"
128
+ },
129
+ {
130
+ "bbox_3d": [
131
+ 2.179,
132
+ 1.598,
133
+ 1.537,
134
+ 0.259,
135
+ 1.411,
136
+ 0.935,
137
+ 0.0,
138
+ 0.0,
139
+ -1.666
140
+ ],
141
+ "label": "door"
142
+ },
143
+ {
144
+ "bbox_3d": [
145
+ 4.541,
146
+ 3.89,
147
+ 1.283,
148
+ 0.607,
149
+ 0.792,
150
+ 0.581,
151
+ 0.0,
152
+ 0.0,
153
+ -1.595
154
+ ],
155
+ "label": "chair"
156
+ },
157
+ {
158
+ "bbox_3d": [
159
+ 6.314,
160
+ 3.931,
161
+ 1.398,
162
+ 0.485,
163
+ 0.304,
164
+ 0.958,
165
+ 0.0,
166
+ 0.0,
167
+ -1.736
168
+ ],
169
+ "label": "window"
170
+ },
171
+ {
172
+ "bbox_3d": [
173
+ 6.047,
174
+ 4.378,
175
+ 1.023,
176
+ 0.827,
177
+ 0.64,
178
+ 1.523,
179
+ 0.0,
180
+ 0.0,
181
+ -1.578
182
+ ],
183
+ "label": "table"
184
+ },
185
+ {
186
+ "bbox_3d": [
187
+ 6.46,
188
+ 5.208,
189
+ 1.005,
190
+ 0.524,
191
+ 0.844,
192
+ 0.445,
193
+ 0.0,
194
+ 0.0,
195
+ -1.44
196
+ ],
197
+ "label": "chair"
198
+ }
199
+ ]
200
+ }
examples/scene0251_00/vqa_input.json ADDED
@@ -0,0 +1,1483 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "id": 3761,
4
+ "dataset": "scannet",
5
+ "scene_name": "scene0251_00",
6
+ "question_type": "room_size_estimation",
7
+ "question": "What is the size of this room (in square meters)? \nIf multiple rooms are shown, estimate the size of the combined space.",
8
+ "ground_truth": "35.9",
9
+ "options": null,
10
+ "key_object": [],
11
+ "qa_type": "room_size_estimation",
12
+ "gt_3d_bboxes": [
13
+ {
14
+ "bbox_3d": [
15
+ 4.428,
16
+ 7.408,
17
+ 1.642,
18
+ 0.455,
19
+ 0.258,
20
+ 0.931,
21
+ 0.0,
22
+ 0.0,
23
+ -1.395
24
+ ],
25
+ "label": "window"
26
+ },
27
+ {
28
+ "bbox_3d": [
29
+ 4.209,
30
+ 6.806,
31
+ 0.962,
32
+ 1.357,
33
+ 0.736,
34
+ 1.514,
35
+ 0.0,
36
+ 0.0,
37
+ -1.492
38
+ ],
39
+ "label": "table"
40
+ },
41
+ {
42
+ "bbox_3d": [
43
+ 3.927,
44
+ 7.236,
45
+ 1.853,
46
+ 0.165,
47
+ 0.575,
48
+ 1.046,
49
+ 0.0,
50
+ 0.0,
51
+ -1.37
52
+ ],
53
+ "label": "window"
54
+ },
55
+ {
56
+ "bbox_3d": [
57
+ 1.194,
58
+ 4.451,
59
+ 1.528,
60
+ 0.459,
61
+ 0.533,
62
+ 0.447,
63
+ 0.0,
64
+ 0.0,
65
+ -1.655
66
+ ],
67
+ "label": "trash bin"
68
+ },
69
+ {
70
+ "bbox_3d": [
71
+ 0.993,
72
+ 3.58,
73
+ 1.528,
74
+ 0.497,
75
+ 0.657,
76
+ 0.434,
77
+ 0.0,
78
+ 0.0,
79
+ -1.484
80
+ ],
81
+ "label": "trash bin"
82
+ },
83
+ {
84
+ "bbox_3d": [
85
+ 0.626,
86
+ 3.272,
87
+ 1.709,
88
+ 0.412,
89
+ 0.619,
90
+ 0.421,
91
+ 0.0,
92
+ 0.0,
93
+ -1.548
94
+ ],
95
+ "label": "trash bin"
96
+ },
97
+ {
98
+ "bbox_3d": [
99
+ 0.654,
100
+ 2.783,
101
+ 1.572,
102
+ 0.519,
103
+ 0.664,
104
+ 0.526,
105
+ 0.0,
106
+ 0.0,
107
+ -1.618
108
+ ],
109
+ "label": "trash bin"
110
+ },
111
+ {
112
+ "bbox_3d": [
113
+ 0.823,
114
+ 2.27,
115
+ 1.4,
116
+ 0.52,
117
+ 0.608,
118
+ 0.564,
119
+ 0.0,
120
+ 0.0,
121
+ -1.619
122
+ ],
123
+ "label": "trash bin"
124
+ },
125
+ {
126
+ "bbox_3d": [
127
+ 1.16,
128
+ 1.304,
129
+ 1.655,
130
+ 0.222,
131
+ 1.298,
132
+ 0.781,
133
+ 0.0,
134
+ 0.0,
135
+ -1.693
136
+ ],
137
+ "label": "door"
138
+ },
139
+ {
140
+ "bbox_3d": [
141
+ 2.179,
142
+ 1.598,
143
+ 1.537,
144
+ 0.259,
145
+ 1.411,
146
+ 0.935,
147
+ 0.0,
148
+ 0.0,
149
+ -1.666
150
+ ],
151
+ "label": "door"
152
+ },
153
+ {
154
+ "bbox_3d": [
155
+ 4.541,
156
+ 3.89,
157
+ 1.283,
158
+ 0.607,
159
+ 0.792,
160
+ 0.581,
161
+ 0.0,
162
+ 0.0,
163
+ -1.595
164
+ ],
165
+ "label": "chair"
166
+ },
167
+ {
168
+ "bbox_3d": [
169
+ 6.314,
170
+ 3.931,
171
+ 1.398,
172
+ 0.485,
173
+ 0.304,
174
+ 0.958,
175
+ 0.0,
176
+ 0.0,
177
+ -1.736
178
+ ],
179
+ "label": "window"
180
+ },
181
+ {
182
+ "bbox_3d": [
183
+ 6.047,
184
+ 4.378,
185
+ 1.023,
186
+ 0.827,
187
+ 0.64,
188
+ 1.523,
189
+ 0.0,
190
+ 0.0,
191
+ -1.578
192
+ ],
193
+ "label": "table"
194
+ },
195
+ {
196
+ "bbox_3d": [
197
+ 6.46,
198
+ 5.208,
199
+ 1.005,
200
+ 0.524,
201
+ 0.844,
202
+ 0.445,
203
+ 0.0,
204
+ 0.0,
205
+ -1.44
206
+ ],
207
+ "label": "chair"
208
+ }
209
+ ]
210
+ },
211
+ {
212
+ "id": 4386,
213
+ "dataset": "scannet",
214
+ "scene_name": "scene0251_00",
215
+ "question_type": "object_counting",
216
+ "question": "How many chair(s) are in this room?",
217
+ "ground_truth": "2",
218
+ "options": null,
219
+ "key_object": [
220
+ "chair"
221
+ ],
222
+ "qa_type": "object_counting",
223
+ "gt_3d_bboxes": [
224
+ {
225
+ "bbox_3d": [
226
+ 4.428,
227
+ 7.408,
228
+ 1.642,
229
+ 0.455,
230
+ 0.258,
231
+ 0.931,
232
+ 0.0,
233
+ 0.0,
234
+ -1.395
235
+ ],
236
+ "label": "window"
237
+ },
238
+ {
239
+ "bbox_3d": [
240
+ 4.209,
241
+ 6.806,
242
+ 0.962,
243
+ 1.357,
244
+ 0.736,
245
+ 1.514,
246
+ 0.0,
247
+ 0.0,
248
+ -1.492
249
+ ],
250
+ "label": "table"
251
+ },
252
+ {
253
+ "bbox_3d": [
254
+ 3.927,
255
+ 7.236,
256
+ 1.853,
257
+ 0.165,
258
+ 0.575,
259
+ 1.046,
260
+ 0.0,
261
+ 0.0,
262
+ -1.37
263
+ ],
264
+ "label": "window"
265
+ },
266
+ {
267
+ "bbox_3d": [
268
+ 1.194,
269
+ 4.451,
270
+ 1.528,
271
+ 0.459,
272
+ 0.533,
273
+ 0.447,
274
+ 0.0,
275
+ 0.0,
276
+ -1.655
277
+ ],
278
+ "label": "trash bin"
279
+ },
280
+ {
281
+ "bbox_3d": [
282
+ 0.993,
283
+ 3.58,
284
+ 1.528,
285
+ 0.497,
286
+ 0.657,
287
+ 0.434,
288
+ 0.0,
289
+ 0.0,
290
+ -1.484
291
+ ],
292
+ "label": "trash bin"
293
+ },
294
+ {
295
+ "bbox_3d": [
296
+ 0.626,
297
+ 3.272,
298
+ 1.709,
299
+ 0.412,
300
+ 0.619,
301
+ 0.421,
302
+ 0.0,
303
+ 0.0,
304
+ -1.548
305
+ ],
306
+ "label": "trash bin"
307
+ },
308
+ {
309
+ "bbox_3d": [
310
+ 0.654,
311
+ 2.783,
312
+ 1.572,
313
+ 0.519,
314
+ 0.664,
315
+ 0.526,
316
+ 0.0,
317
+ 0.0,
318
+ -1.618
319
+ ],
320
+ "label": "trash bin"
321
+ },
322
+ {
323
+ "bbox_3d": [
324
+ 0.823,
325
+ 2.27,
326
+ 1.4,
327
+ 0.52,
328
+ 0.608,
329
+ 0.564,
330
+ 0.0,
331
+ 0.0,
332
+ -1.619
333
+ ],
334
+ "label": "trash bin"
335
+ },
336
+ {
337
+ "bbox_3d": [
338
+ 1.16,
339
+ 1.304,
340
+ 1.655,
341
+ 0.222,
342
+ 1.298,
343
+ 0.781,
344
+ 0.0,
345
+ 0.0,
346
+ -1.693
347
+ ],
348
+ "label": "door"
349
+ },
350
+ {
351
+ "bbox_3d": [
352
+ 2.179,
353
+ 1.598,
354
+ 1.537,
355
+ 0.259,
356
+ 1.411,
357
+ 0.935,
358
+ 0.0,
359
+ 0.0,
360
+ -1.666
361
+ ],
362
+ "label": "door"
363
+ },
364
+ {
365
+ "bbox_3d": [
366
+ 4.541,
367
+ 3.89,
368
+ 1.283,
369
+ 0.607,
370
+ 0.792,
371
+ 0.581,
372
+ 0.0,
373
+ 0.0,
374
+ -1.595
375
+ ],
376
+ "label": "chair"
377
+ },
378
+ {
379
+ "bbox_3d": [
380
+ 6.314,
381
+ 3.931,
382
+ 1.398,
383
+ 0.485,
384
+ 0.304,
385
+ 0.958,
386
+ 0.0,
387
+ 0.0,
388
+ -1.736
389
+ ],
390
+ "label": "window"
391
+ },
392
+ {
393
+ "bbox_3d": [
394
+ 6.047,
395
+ 4.378,
396
+ 1.023,
397
+ 0.827,
398
+ 0.64,
399
+ 1.523,
400
+ 0.0,
401
+ 0.0,
402
+ -1.578
403
+ ],
404
+ "label": "table"
405
+ },
406
+ {
407
+ "bbox_3d": [
408
+ 6.46,
409
+ 5.208,
410
+ 1.005,
411
+ 0.524,
412
+ 0.844,
413
+ 0.445,
414
+ 0.0,
415
+ 0.0,
416
+ -1.44
417
+ ],
418
+ "label": "chair"
419
+ }
420
+ ]
421
+ },
422
+ {
423
+ "id": 4387,
424
+ "dataset": "scannet",
425
+ "scene_name": "scene0251_00",
426
+ "question_type": "object_counting",
427
+ "question": "How many window(s) are in this room?",
428
+ "ground_truth": "3",
429
+ "options": null,
430
+ "key_object": [
431
+ "window"
432
+ ],
433
+ "qa_type": "object_counting",
434
+ "gt_3d_bboxes": [
435
+ {
436
+ "bbox_3d": [
437
+ 4.428,
438
+ 7.408,
439
+ 1.642,
440
+ 0.455,
441
+ 0.258,
442
+ 0.931,
443
+ 0.0,
444
+ 0.0,
445
+ -1.395
446
+ ],
447
+ "label": "window"
448
+ },
449
+ {
450
+ "bbox_3d": [
451
+ 4.209,
452
+ 6.806,
453
+ 0.962,
454
+ 1.357,
455
+ 0.736,
456
+ 1.514,
457
+ 0.0,
458
+ 0.0,
459
+ -1.492
460
+ ],
461
+ "label": "table"
462
+ },
463
+ {
464
+ "bbox_3d": [
465
+ 3.927,
466
+ 7.236,
467
+ 1.853,
468
+ 0.165,
469
+ 0.575,
470
+ 1.046,
471
+ 0.0,
472
+ 0.0,
473
+ -1.37
474
+ ],
475
+ "label": "window"
476
+ },
477
+ {
478
+ "bbox_3d": [
479
+ 1.194,
480
+ 4.451,
481
+ 1.528,
482
+ 0.459,
483
+ 0.533,
484
+ 0.447,
485
+ 0.0,
486
+ 0.0,
487
+ -1.655
488
+ ],
489
+ "label": "trash bin"
490
+ },
491
+ {
492
+ "bbox_3d": [
493
+ 0.993,
494
+ 3.58,
495
+ 1.528,
496
+ 0.497,
497
+ 0.657,
498
+ 0.434,
499
+ 0.0,
500
+ 0.0,
501
+ -1.484
502
+ ],
503
+ "label": "trash bin"
504
+ },
505
+ {
506
+ "bbox_3d": [
507
+ 0.626,
508
+ 3.272,
509
+ 1.709,
510
+ 0.412,
511
+ 0.619,
512
+ 0.421,
513
+ 0.0,
514
+ 0.0,
515
+ -1.548
516
+ ],
517
+ "label": "trash bin"
518
+ },
519
+ {
520
+ "bbox_3d": [
521
+ 0.654,
522
+ 2.783,
523
+ 1.572,
524
+ 0.519,
525
+ 0.664,
526
+ 0.526,
527
+ 0.0,
528
+ 0.0,
529
+ -1.618
530
+ ],
531
+ "label": "trash bin"
532
+ },
533
+ {
534
+ "bbox_3d": [
535
+ 0.823,
536
+ 2.27,
537
+ 1.4,
538
+ 0.52,
539
+ 0.608,
540
+ 0.564,
541
+ 0.0,
542
+ 0.0,
543
+ -1.619
544
+ ],
545
+ "label": "trash bin"
546
+ },
547
+ {
548
+ "bbox_3d": [
549
+ 1.16,
550
+ 1.304,
551
+ 1.655,
552
+ 0.222,
553
+ 1.298,
554
+ 0.781,
555
+ 0.0,
556
+ 0.0,
557
+ -1.693
558
+ ],
559
+ "label": "door"
560
+ },
561
+ {
562
+ "bbox_3d": [
563
+ 2.179,
564
+ 1.598,
565
+ 1.537,
566
+ 0.259,
567
+ 1.411,
568
+ 0.935,
569
+ 0.0,
570
+ 0.0,
571
+ -1.666
572
+ ],
573
+ "label": "door"
574
+ },
575
+ {
576
+ "bbox_3d": [
577
+ 4.541,
578
+ 3.89,
579
+ 1.283,
580
+ 0.607,
581
+ 0.792,
582
+ 0.581,
583
+ 0.0,
584
+ 0.0,
585
+ -1.595
586
+ ],
587
+ "label": "chair"
588
+ },
589
+ {
590
+ "bbox_3d": [
591
+ 6.314,
592
+ 3.931,
593
+ 1.398,
594
+ 0.485,
595
+ 0.304,
596
+ 0.958,
597
+ 0.0,
598
+ 0.0,
599
+ -1.736
600
+ ],
601
+ "label": "window"
602
+ },
603
+ {
604
+ "bbox_3d": [
605
+ 6.047,
606
+ 4.378,
607
+ 1.023,
608
+ 0.827,
609
+ 0.64,
610
+ 1.523,
611
+ 0.0,
612
+ 0.0,
613
+ -1.578
614
+ ],
615
+ "label": "table"
616
+ },
617
+ {
618
+ "bbox_3d": [
619
+ 6.46,
620
+ 5.208,
621
+ 1.005,
622
+ 0.524,
623
+ 0.844,
624
+ 0.445,
625
+ 0.0,
626
+ 0.0,
627
+ -1.44
628
+ ],
629
+ "label": "chair"
630
+ }
631
+ ]
632
+ },
633
+ {
634
+ "id": 4388,
635
+ "dataset": "scannet",
636
+ "scene_name": "scene0251_00",
637
+ "question_type": "object_counting",
638
+ "question": "How many door(s) are in this room?",
639
+ "ground_truth": "3",
640
+ "options": null,
641
+ "key_object": [
642
+ "door"
643
+ ],
644
+ "qa_type": "object_counting",
645
+ "gt_3d_bboxes": [
646
+ {
647
+ "bbox_3d": [
648
+ 4.428,
649
+ 7.408,
650
+ 1.642,
651
+ 0.455,
652
+ 0.258,
653
+ 0.931,
654
+ 0.0,
655
+ 0.0,
656
+ -1.395
657
+ ],
658
+ "label": "window"
659
+ },
660
+ {
661
+ "bbox_3d": [
662
+ 4.209,
663
+ 6.806,
664
+ 0.962,
665
+ 1.357,
666
+ 0.736,
667
+ 1.514,
668
+ 0.0,
669
+ 0.0,
670
+ -1.492
671
+ ],
672
+ "label": "table"
673
+ },
674
+ {
675
+ "bbox_3d": [
676
+ 3.927,
677
+ 7.236,
678
+ 1.853,
679
+ 0.165,
680
+ 0.575,
681
+ 1.046,
682
+ 0.0,
683
+ 0.0,
684
+ -1.37
685
+ ],
686
+ "label": "window"
687
+ },
688
+ {
689
+ "bbox_3d": [
690
+ 1.194,
691
+ 4.451,
692
+ 1.528,
693
+ 0.459,
694
+ 0.533,
695
+ 0.447,
696
+ 0.0,
697
+ 0.0,
698
+ -1.655
699
+ ],
700
+ "label": "trash bin"
701
+ },
702
+ {
703
+ "bbox_3d": [
704
+ 0.993,
705
+ 3.58,
706
+ 1.528,
707
+ 0.497,
708
+ 0.657,
709
+ 0.434,
710
+ 0.0,
711
+ 0.0,
712
+ -1.484
713
+ ],
714
+ "label": "trash bin"
715
+ },
716
+ {
717
+ "bbox_3d": [
718
+ 0.626,
719
+ 3.272,
720
+ 1.709,
721
+ 0.412,
722
+ 0.619,
723
+ 0.421,
724
+ 0.0,
725
+ 0.0,
726
+ -1.548
727
+ ],
728
+ "label": "trash bin"
729
+ },
730
+ {
731
+ "bbox_3d": [
732
+ 0.654,
733
+ 2.783,
734
+ 1.572,
735
+ 0.519,
736
+ 0.664,
737
+ 0.526,
738
+ 0.0,
739
+ 0.0,
740
+ -1.618
741
+ ],
742
+ "label": "trash bin"
743
+ },
744
+ {
745
+ "bbox_3d": [
746
+ 0.823,
747
+ 2.27,
748
+ 1.4,
749
+ 0.52,
750
+ 0.608,
751
+ 0.564,
752
+ 0.0,
753
+ 0.0,
754
+ -1.619
755
+ ],
756
+ "label": "trash bin"
757
+ },
758
+ {
759
+ "bbox_3d": [
760
+ 1.16,
761
+ 1.304,
762
+ 1.655,
763
+ 0.222,
764
+ 1.298,
765
+ 0.781,
766
+ 0.0,
767
+ 0.0,
768
+ -1.693
769
+ ],
770
+ "label": "door"
771
+ },
772
+ {
773
+ "bbox_3d": [
774
+ 2.179,
775
+ 1.598,
776
+ 1.537,
777
+ 0.259,
778
+ 1.411,
779
+ 0.935,
780
+ 0.0,
781
+ 0.0,
782
+ -1.666
783
+ ],
784
+ "label": "door"
785
+ },
786
+ {
787
+ "bbox_3d": [
788
+ 4.541,
789
+ 3.89,
790
+ 1.283,
791
+ 0.607,
792
+ 0.792,
793
+ 0.581,
794
+ 0.0,
795
+ 0.0,
796
+ -1.595
797
+ ],
798
+ "label": "chair"
799
+ },
800
+ {
801
+ "bbox_3d": [
802
+ 6.314,
803
+ 3.931,
804
+ 1.398,
805
+ 0.485,
806
+ 0.304,
807
+ 0.958,
808
+ 0.0,
809
+ 0.0,
810
+ -1.736
811
+ ],
812
+ "label": "window"
813
+ },
814
+ {
815
+ "bbox_3d": [
816
+ 6.047,
817
+ 4.378,
818
+ 1.023,
819
+ 0.827,
820
+ 0.64,
821
+ 1.523,
822
+ 0.0,
823
+ 0.0,
824
+ -1.578
825
+ ],
826
+ "label": "table"
827
+ },
828
+ {
829
+ "bbox_3d": [
830
+ 6.46,
831
+ 5.208,
832
+ 1.005,
833
+ 0.524,
834
+ 0.844,
835
+ 0.445,
836
+ 0.0,
837
+ 0.0,
838
+ -1.44
839
+ ],
840
+ "label": "chair"
841
+ }
842
+ ]
843
+ },
844
+ {
845
+ "id": 4389,
846
+ "dataset": "scannet",
847
+ "scene_name": "scene0251_00",
848
+ "question_type": "object_counting",
849
+ "question": "How many trash bin(s) are in this room?",
850
+ "ground_truth": "5",
851
+ "options": null,
852
+ "key_object": [
853
+ "trash bin"
854
+ ],
855
+ "qa_type": "object_counting",
856
+ "gt_3d_bboxes": [
857
+ {
858
+ "bbox_3d": [
859
+ 4.428,
860
+ 7.408,
861
+ 1.642,
862
+ 0.455,
863
+ 0.258,
864
+ 0.931,
865
+ 0.0,
866
+ 0.0,
867
+ -1.395
868
+ ],
869
+ "label": "window"
870
+ },
871
+ {
872
+ "bbox_3d": [
873
+ 4.209,
874
+ 6.806,
875
+ 0.962,
876
+ 1.357,
877
+ 0.736,
878
+ 1.514,
879
+ 0.0,
880
+ 0.0,
881
+ -1.492
882
+ ],
883
+ "label": "table"
884
+ },
885
+ {
886
+ "bbox_3d": [
887
+ 3.927,
888
+ 7.236,
889
+ 1.853,
890
+ 0.165,
891
+ 0.575,
892
+ 1.046,
893
+ 0.0,
894
+ 0.0,
895
+ -1.37
896
+ ],
897
+ "label": "window"
898
+ },
899
+ {
900
+ "bbox_3d": [
901
+ 1.194,
902
+ 4.451,
903
+ 1.528,
904
+ 0.459,
905
+ 0.533,
906
+ 0.447,
907
+ 0.0,
908
+ 0.0,
909
+ -1.655
910
+ ],
911
+ "label": "trash bin"
912
+ },
913
+ {
914
+ "bbox_3d": [
915
+ 0.993,
916
+ 3.58,
917
+ 1.528,
918
+ 0.497,
919
+ 0.657,
920
+ 0.434,
921
+ 0.0,
922
+ 0.0,
923
+ -1.484
924
+ ],
925
+ "label": "trash bin"
926
+ },
927
+ {
928
+ "bbox_3d": [
929
+ 0.626,
930
+ 3.272,
931
+ 1.709,
932
+ 0.412,
933
+ 0.619,
934
+ 0.421,
935
+ 0.0,
936
+ 0.0,
937
+ -1.548
938
+ ],
939
+ "label": "trash bin"
940
+ },
941
+ {
942
+ "bbox_3d": [
943
+ 0.654,
944
+ 2.783,
945
+ 1.572,
946
+ 0.519,
947
+ 0.664,
948
+ 0.526,
949
+ 0.0,
950
+ 0.0,
951
+ -1.618
952
+ ],
953
+ "label": "trash bin"
954
+ },
955
+ {
956
+ "bbox_3d": [
957
+ 0.823,
958
+ 2.27,
959
+ 1.4,
960
+ 0.52,
961
+ 0.608,
962
+ 0.564,
963
+ 0.0,
964
+ 0.0,
965
+ -1.619
966
+ ],
967
+ "label": "trash bin"
968
+ },
969
+ {
970
+ "bbox_3d": [
971
+ 1.16,
972
+ 1.304,
973
+ 1.655,
974
+ 0.222,
975
+ 1.298,
976
+ 0.781,
977
+ 0.0,
978
+ 0.0,
979
+ -1.693
980
+ ],
981
+ "label": "door"
982
+ },
983
+ {
984
+ "bbox_3d": [
985
+ 2.179,
986
+ 1.598,
987
+ 1.537,
988
+ 0.259,
989
+ 1.411,
990
+ 0.935,
991
+ 0.0,
992
+ 0.0,
993
+ -1.666
994
+ ],
995
+ "label": "door"
996
+ },
997
+ {
998
+ "bbox_3d": [
999
+ 4.541,
1000
+ 3.89,
1001
+ 1.283,
1002
+ 0.607,
1003
+ 0.792,
1004
+ 0.581,
1005
+ 0.0,
1006
+ 0.0,
1007
+ -1.595
1008
+ ],
1009
+ "label": "chair"
1010
+ },
1011
+ {
1012
+ "bbox_3d": [
1013
+ 6.314,
1014
+ 3.931,
1015
+ 1.398,
1016
+ 0.485,
1017
+ 0.304,
1018
+ 0.958,
1019
+ 0.0,
1020
+ 0.0,
1021
+ -1.736
1022
+ ],
1023
+ "label": "window"
1024
+ },
1025
+ {
1026
+ "bbox_3d": [
1027
+ 6.047,
1028
+ 4.378,
1029
+ 1.023,
1030
+ 0.827,
1031
+ 0.64,
1032
+ 1.523,
1033
+ 0.0,
1034
+ 0.0,
1035
+ -1.578
1036
+ ],
1037
+ "label": "table"
1038
+ },
1039
+ {
1040
+ "bbox_3d": [
1041
+ 6.46,
1042
+ 5.208,
1043
+ 1.005,
1044
+ 0.524,
1045
+ 0.844,
1046
+ 0.445,
1047
+ 0.0,
1048
+ 0.0,
1049
+ -1.44
1050
+ ],
1051
+ "label": "chair"
1052
+ }
1053
+ ]
1054
+ },
1055
+ {
1056
+ "id": 4390,
1057
+ "dataset": "scannet",
1058
+ "scene_name": "scene0251_00",
1059
+ "question_type": "object_counting",
1060
+ "question": "How many table(s) are in this room?",
1061
+ "ground_truth": "2",
1062
+ "options": null,
1063
+ "key_object": [
1064
+ "table"
1065
+ ],
1066
+ "qa_type": "object_counting",
1067
+ "gt_3d_bboxes": [
1068
+ {
1069
+ "bbox_3d": [
1070
+ 4.428,
1071
+ 7.408,
1072
+ 1.642,
1073
+ 0.455,
1074
+ 0.258,
1075
+ 0.931,
1076
+ 0.0,
1077
+ 0.0,
1078
+ -1.395
1079
+ ],
1080
+ "label": "window"
1081
+ },
1082
+ {
1083
+ "bbox_3d": [
1084
+ 4.209,
1085
+ 6.806,
1086
+ 0.962,
1087
+ 1.357,
1088
+ 0.736,
1089
+ 1.514,
1090
+ 0.0,
1091
+ 0.0,
1092
+ -1.492
1093
+ ],
1094
+ "label": "table"
1095
+ },
1096
+ {
1097
+ "bbox_3d": [
1098
+ 3.927,
1099
+ 7.236,
1100
+ 1.853,
1101
+ 0.165,
1102
+ 0.575,
1103
+ 1.046,
1104
+ 0.0,
1105
+ 0.0,
1106
+ -1.37
1107
+ ],
1108
+ "label": "window"
1109
+ },
1110
+ {
1111
+ "bbox_3d": [
1112
+ 1.194,
1113
+ 4.451,
1114
+ 1.528,
1115
+ 0.459,
1116
+ 0.533,
1117
+ 0.447,
1118
+ 0.0,
1119
+ 0.0,
1120
+ -1.655
1121
+ ],
1122
+ "label": "trash bin"
1123
+ },
1124
+ {
1125
+ "bbox_3d": [
1126
+ 0.993,
1127
+ 3.58,
1128
+ 1.528,
1129
+ 0.497,
1130
+ 0.657,
1131
+ 0.434,
1132
+ 0.0,
1133
+ 0.0,
1134
+ -1.484
1135
+ ],
1136
+ "label": "trash bin"
1137
+ },
1138
+ {
1139
+ "bbox_3d": [
1140
+ 0.626,
1141
+ 3.272,
1142
+ 1.709,
1143
+ 0.412,
1144
+ 0.619,
1145
+ 0.421,
1146
+ 0.0,
1147
+ 0.0,
1148
+ -1.548
1149
+ ],
1150
+ "label": "trash bin"
1151
+ },
1152
+ {
1153
+ "bbox_3d": [
1154
+ 0.654,
1155
+ 2.783,
1156
+ 1.572,
1157
+ 0.519,
1158
+ 0.664,
1159
+ 0.526,
1160
+ 0.0,
1161
+ 0.0,
1162
+ -1.618
1163
+ ],
1164
+ "label": "trash bin"
1165
+ },
1166
+ {
1167
+ "bbox_3d": [
1168
+ 0.823,
1169
+ 2.27,
1170
+ 1.4,
1171
+ 0.52,
1172
+ 0.608,
1173
+ 0.564,
1174
+ 0.0,
1175
+ 0.0,
1176
+ -1.619
1177
+ ],
1178
+ "label": "trash bin"
1179
+ },
1180
+ {
1181
+ "bbox_3d": [
1182
+ 1.16,
1183
+ 1.304,
1184
+ 1.655,
1185
+ 0.222,
1186
+ 1.298,
1187
+ 0.781,
1188
+ 0.0,
1189
+ 0.0,
1190
+ -1.693
1191
+ ],
1192
+ "label": "door"
1193
+ },
1194
+ {
1195
+ "bbox_3d": [
1196
+ 2.179,
1197
+ 1.598,
1198
+ 1.537,
1199
+ 0.259,
1200
+ 1.411,
1201
+ 0.935,
1202
+ 0.0,
1203
+ 0.0,
1204
+ -1.666
1205
+ ],
1206
+ "label": "door"
1207
+ },
1208
+ {
1209
+ "bbox_3d": [
1210
+ 4.541,
1211
+ 3.89,
1212
+ 1.283,
1213
+ 0.607,
1214
+ 0.792,
1215
+ 0.581,
1216
+ 0.0,
1217
+ 0.0,
1218
+ -1.595
1219
+ ],
1220
+ "label": "chair"
1221
+ },
1222
+ {
1223
+ "bbox_3d": [
1224
+ 6.314,
1225
+ 3.931,
1226
+ 1.398,
1227
+ 0.485,
1228
+ 0.304,
1229
+ 0.958,
1230
+ 0.0,
1231
+ 0.0,
1232
+ -1.736
1233
+ ],
1234
+ "label": "window"
1235
+ },
1236
+ {
1237
+ "bbox_3d": [
1238
+ 6.047,
1239
+ 4.378,
1240
+ 1.023,
1241
+ 0.827,
1242
+ 0.64,
1243
+ 1.523,
1244
+ 0.0,
1245
+ 0.0,
1246
+ -1.578
1247
+ ],
1248
+ "label": "table"
1249
+ },
1250
+ {
1251
+ "bbox_3d": [
1252
+ 6.46,
1253
+ 5.208,
1254
+ 1.005,
1255
+ 0.524,
1256
+ 0.844,
1257
+ 0.445,
1258
+ 0.0,
1259
+ 0.0,
1260
+ -1.44
1261
+ ],
1262
+ "label": "chair"
1263
+ }
1264
+ ]
1265
+ },
1266
+ {
1267
+ "id": 4994,
1268
+ "dataset": "scannet",
1269
+ "scene_name": "scene0251_00",
1270
+ "question_type": "route_planning",
1271
+ "question": "You are a robot beginning at the poster and facing the cabinet. You want to navigate to the trashbin. You will perform the following actions (Note: for each [please fill in], choose either 'turn back,' 'turn left,' or 'turn right.'): 1. Go forward until the cabinet. 2. [please fill in] 3. Go forward until the trashbin. You have reached the final destination.",
1272
+ "ground_truth": "A",
1273
+ "options": [
1274
+ "A. Turn Left",
1275
+ "B. Turn Right",
1276
+ "C. Turn Back"
1277
+ ],
1278
+ "key_object": [
1279
+ "poster",
1280
+ "cabinet",
1281
+ "trashbin"
1282
+ ],
1283
+ "qa_type": "route_planning",
1284
+ "gt_3d_bboxes": [
1285
+ {
1286
+ "bbox_3d": [
1287
+ 4.428,
1288
+ 7.408,
1289
+ 1.642,
1290
+ 0.455,
1291
+ 0.258,
1292
+ 0.931,
1293
+ 0.0,
1294
+ 0.0,
1295
+ -1.395
1296
+ ],
1297
+ "label": "window"
1298
+ },
1299
+ {
1300
+ "bbox_3d": [
1301
+ 4.209,
1302
+ 6.806,
1303
+ 0.962,
1304
+ 1.357,
1305
+ 0.736,
1306
+ 1.514,
1307
+ 0.0,
1308
+ 0.0,
1309
+ -1.492
1310
+ ],
1311
+ "label": "table"
1312
+ },
1313
+ {
1314
+ "bbox_3d": [
1315
+ 3.927,
1316
+ 7.236,
1317
+ 1.853,
1318
+ 0.165,
1319
+ 0.575,
1320
+ 1.046,
1321
+ 0.0,
1322
+ 0.0,
1323
+ -1.37
1324
+ ],
1325
+ "label": "window"
1326
+ },
1327
+ {
1328
+ "bbox_3d": [
1329
+ 1.194,
1330
+ 4.451,
1331
+ 1.528,
1332
+ 0.459,
1333
+ 0.533,
1334
+ 0.447,
1335
+ 0.0,
1336
+ 0.0,
1337
+ -1.655
1338
+ ],
1339
+ "label": "trash bin"
1340
+ },
1341
+ {
1342
+ "bbox_3d": [
1343
+ 0.993,
1344
+ 3.58,
1345
+ 1.528,
1346
+ 0.497,
1347
+ 0.657,
1348
+ 0.434,
1349
+ 0.0,
1350
+ 0.0,
1351
+ -1.484
1352
+ ],
1353
+ "label": "trash bin"
1354
+ },
1355
+ {
1356
+ "bbox_3d": [
1357
+ 0.626,
1358
+ 3.272,
1359
+ 1.709,
1360
+ 0.412,
1361
+ 0.619,
1362
+ 0.421,
1363
+ 0.0,
1364
+ 0.0,
1365
+ -1.548
1366
+ ],
1367
+ "label": "trash bin"
1368
+ },
1369
+ {
1370
+ "bbox_3d": [
1371
+ 0.654,
1372
+ 2.783,
1373
+ 1.572,
1374
+ 0.519,
1375
+ 0.664,
1376
+ 0.526,
1377
+ 0.0,
1378
+ 0.0,
1379
+ -1.618
1380
+ ],
1381
+ "label": "trash bin"
1382
+ },
1383
+ {
1384
+ "bbox_3d": [
1385
+ 0.823,
1386
+ 2.27,
1387
+ 1.4,
1388
+ 0.52,
1389
+ 0.608,
1390
+ 0.564,
1391
+ 0.0,
1392
+ 0.0,
1393
+ -1.619
1394
+ ],
1395
+ "label": "trash bin"
1396
+ },
1397
+ {
1398
+ "bbox_3d": [
1399
+ 1.16,
1400
+ 1.304,
1401
+ 1.655,
1402
+ 0.222,
1403
+ 1.298,
1404
+ 0.781,
1405
+ 0.0,
1406
+ 0.0,
1407
+ -1.693
1408
+ ],
1409
+ "label": "door"
1410
+ },
1411
+ {
1412
+ "bbox_3d": [
1413
+ 2.179,
1414
+ 1.598,
1415
+ 1.537,
1416
+ 0.259,
1417
+ 1.411,
1418
+ 0.935,
1419
+ 0.0,
1420
+ 0.0,
1421
+ -1.666
1422
+ ],
1423
+ "label": "door"
1424
+ },
1425
+ {
1426
+ "bbox_3d": [
1427
+ 4.541,
1428
+ 3.89,
1429
+ 1.283,
1430
+ 0.607,
1431
+ 0.792,
1432
+ 0.581,
1433
+ 0.0,
1434
+ 0.0,
1435
+ -1.595
1436
+ ],
1437
+ "label": "chair"
1438
+ },
1439
+ {
1440
+ "bbox_3d": [
1441
+ 6.314,
1442
+ 3.931,
1443
+ 1.398,
1444
+ 0.485,
1445
+ 0.304,
1446
+ 0.958,
1447
+ 0.0,
1448
+ 0.0,
1449
+ -1.736
1450
+ ],
1451
+ "label": "window"
1452
+ },
1453
+ {
1454
+ "bbox_3d": [
1455
+ 6.047,
1456
+ 4.378,
1457
+ 1.023,
1458
+ 0.827,
1459
+ 0.64,
1460
+ 1.523,
1461
+ 0.0,
1462
+ 0.0,
1463
+ -1.578
1464
+ ],
1465
+ "label": "table"
1466
+ },
1467
+ {
1468
+ "bbox_3d": [
1469
+ 6.46,
1470
+ 5.208,
1471
+ 1.005,
1472
+ 0.524,
1473
+ 0.844,
1474
+ 0.445,
1475
+ 0.0,
1476
+ 0.0,
1477
+ -1.44
1478
+ ],
1479
+ "label": "chair"
1480
+ }
1481
+ ]
1482
+ }
1483
+ ]