Dataset Viewer
Auto-converted to Parquet
id
int64
0
7.21k
img_type
stringclasses
2 values
format_type
stringclasses
2 values
task
stringclasses
20 values
source
stringclasses
2 values
image
images listlengths
1
3
question
stringlengths
181
1.38k
answer
stringclasses
475 values
0
single_view
fill
depth_prediction_oc
scannet
The camera coordinates show the towel (red point) at 1.0 meters depth. What is the depth of the towel (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
0.9
1
single_view
fill
depth_prediction_oc
scannet
If the center of the chair (red point) is 3.7 meters deep, what is the depth of chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
3.5
2
single_view
fill
depth_prediction_oc
scannet
Knowing that the center of chair (red point) is 1.8 meters deep, estimate the depth of bed (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Your input must be limited to a single number.
1.6
3
single_view
fill
depth_prediction_oc
scannet
Given that toilet paper (red point) is at 1.6 meters, predict the depth of toilet (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
1.5
4
single_view
fill
depth_prediction_oc
scannet
The dresser (red point) is placed at a depth of 3.0 meters along the Z-axis(camera coordinate system). Estimate the depth of laundry basket (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Please give a response in the form of a single number.
2.1
5
single_view
fill
depth_prediction_oc
scannet
Using the chair (red point) depth of 3.4 meters, determine the depth of tv (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter one numeric value as your answer.
3.0
6
single_view
fill
depth_prediction_oc
scannet
The depth of the cabinet (red point) at its center is measured as 1.6 meters. Estimate the depth of the computer tower (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
1.4
7
single_view
fill
depth_prediction_oc
scannet
The range hood (red point) is placed at a depth of 1.8 meters along the Z-axis(camera coordinate system). Estimate the depth of picture (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
1.9
8
single_view
fill
depth_prediction_oc
scannet
Using sink (red point) at 1.6 meters as a guide, determine the depth of trash can (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Respond with exactly one numeric value.
1.9
9
single_view
fill
depth_prediction_oc
scannet
With trash can (red point) at 1.6 meters, calculate the depth of toilet (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer by providing just one numeric value.
1.3
10
single_view
fill
depth_prediction_oc
scannet
The monitor (red point) is positioned at 1.6 meters depth. Predict the depth of the keyboard (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide a numeric response with just one value.
1.5
11
single_view
fill
depth_prediction_oc
scannet
If the pillow (red point) is positioned at 2.3 meters, estimate the depth of the armchair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Type in exactly one number as your reply.
3.1
12
single_view
fill
depth_prediction_oc
scannet
The depth of the towel (red point) is given as 1.2 meters. Use this to estimate the depth of toilet paper (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
1.8
13
single_view
fill
depth_prediction_oc
scannet
If the box (red point) is positioned at 1.6 meters, estimate the depth of the box (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Please give a response in the form of a single number.
1.7
14
single_view
fill
depth_prediction_oc
scannet
The shower curtain (red point) lies at 1.3 meters depth. Can you estimate the depth of toilet paper holder (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
2.8
15
single_view
fill
depth_prediction_oc
scannet
Knowing the depth of the sink (red point) at 1.1 meters, predict the depth of the bag (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.0
16
single_view
fill
depth_prediction_oc
scannet
Knowing the depth of the column (red point) at 2.9 meters, predict the depth of the chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Respond with exactly one numeric value.
3.5
17
single_view
fill
depth_prediction_oc
scannet
If the window (red point) is 3.1 meters deep, how deep is the fire extinguisher (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
2.9
18
single_view
fill
depth_prediction_oc
scannet
With box (red point) at 3.7 meters, calculate the depth of cart (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
5.4
19
single_view
fill
depth_prediction_oc
scannet
With the picture (red point) depth measured at 2.4 meters, determine the depth of picture (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
3.4
20
single_view
fill
depth_prediction_oc
scannet
If stool (red point) has a depth of 2.2 meters, how deep is the curtain (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide a numeric response with just one value.
2.1
21
single_view
fill
depth_prediction_oc
scannet
The bottle (red point) is positioned at 1.2 meters depth. Predict the depth of the chair (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
1.6
22
single_view
fill
depth_prediction_oc
scannet
The computer tower (red point) is 2.0 meters deep in the camera coordinate system. Predict the depth of the chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one number should be included in your response.
2.2
23
single_view
fill
depth_prediction_oc
scannet
If the center of the cabinet (red point) is 5.1 meters deep, what is the depth of shelf (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
3.0
24
single_view
fill
depth_prediction_oc
scannet
Based on the fact that the chair (red point) has a depth of 2.5 meters, what is the depth of the window (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
6.8
25
single_view
fill
depth_prediction_oc
scannet
If the table (red point) is positioned at 2.6 meters, estimate the depth of the clothes (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
1.5
26
single_view
fill
depth_prediction_oc
scannet
If paper bag (red point) has a depth of 1.2 meters, how deep is the bag (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
1.4
27
single_view
fill
depth_prediction_oc
scannet
With the trash can (red point) having a depth of 3.4 meters, determine the depth of desk (blue point) in the same frame. Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer by providing just one numeric value.
1.5
28
single_view
fill
depth_prediction_oc
scannet
The trash bin (red point) lies at 1.4 meters depth. Can you estimate the depth of shelf (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
1.5
29
single_view
fill
depth_prediction_oc
scannet
With table (red point) at a depth of 2.8 meters, determine the approximate depth of shelf (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
2.5
30
single_view
fill
depth_prediction_oc
scannet
Knowing telephone (red point) is at 1.9 meters, calculate the depth of office chair (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter one numeric value as your answer.
1.3
31
single_view
fill
depth_prediction_oc
scannet
With the toilet (red point) depth measured at 0.9 meters, determine the depth of toilet paper (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
1.0
32
single_view
fill
depth_prediction_oc
scannet
The box (red point) is situated at 1.9 meters in the camera's Z-axis(camera coordinate system). What is the depth of the box (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
2.1
33
single_view
fill
depth_prediction_oc
scannet
Knowing that chair (red point) is located at 1.5 meters, what is the depth of refrigerator (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
5.4
34
single_view
fill
depth_prediction_oc
scannet
Given trash can (red point) is positioned at 1.6 meters, how deep is the sink (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Ensure your answer contains only one number.
1.2
35
single_view
fill
depth_prediction_oc
scannet
Given that chair (red point) is at 4.9 meters, predict the depth of chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
4.3
36
single_view
fill
depth_prediction_oc
scannet
With the window (red point) at 1.4 meters, determine the depth of table (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
1.4
37
single_view
fill
depth_prediction_oc
scannet
The chair (red point) lies at a depth of 2.7 meters. How deep is the center of chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
2.1
38
single_view
fill
depth_prediction_oc
scannet
The camera coordinate system positions chair (red point) at a depth of 1.8 meters. Predict the depth of chair (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
2.3
39
single_view
fill
depth_prediction_oc
scannet
Knowing that the center of pillow (red point) is 2.0 meters deep, estimate the depth of pillow (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer by providing just one numeric value.
2.3
40
single_view
fill
depth_prediction_oc
scannet
Given toilet (red point) is positioned at 2.0 meters, how deep is the sink (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Respond with exactly one numeric value.
1.0
41
single_view
fill
depth_prediction_oc
scannet
The camera coordinate system positions armchair (red point) at a depth of 1.3 meters. Predict the depth of armchair (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one number should be included in your response.
2.7
42
single_view
fill
depth_prediction_oc
scannet
Using the book (red point) depth of 1.9 meters, determine the depth of book (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one numeric value is required for your response.
2.2
43
single_view
fill
depth_prediction_oc
scannet
The center of the recycling bin (red point) is positioned at 2.3 meters. Based on this, predict the depth of the center of the stove (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Your input must be limited to a single number.
3.6
44
single_view
fill
depth_prediction_oc
scannet
The monitor (red point) lies at 2.1 meters depth. Predict the depth of the office chair (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
2.6
45
single_view
fill
depth_prediction_oc
scannet
Knowing the depth of the shower walls (red point) at 1.8 meters, predict the depth of the towel (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.3
46
single_view
fill
depth_prediction_oc
scannet
Knowing that laundry detergent (red point) is located at 1.3 meters, what is the depth of shelf (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Respond with exactly one numeric value.
1.3
47
single_view
fill
depth_prediction_oc
scannet
Using person (red point) at 1.1 meters as a guide, determine the depth of monitor (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
2.6
48
single_view
fill
depth_prediction_oc
scannet
With the couch (red point) center located at 2.2 meters in the camera Z-axis(camera coordinate system), determine the depth of backpack (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
2.6
49
single_view
fill
depth_prediction_oc
scannet
The keyboard (red point) at 2.4 meters serves as a depth marker. What is the depth of the computer tower (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Please give a response in the form of a single number.
2.1
50
single_view
fill
depth_prediction_oc
scannet
The cup (red point) at a depth of 1.4 meters serves as a reference. How deep is the picture (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer by providing just one numeric value.
1.9
51
single_view
fill
depth_prediction_oc
scannet
Given that towel (red point) is at 1.0 meters, predict the depth of trash can (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.5
52
single_view
fill
depth_prediction_oc
scannet
Based on the reference depth of vending machine (red point) at 3.3 meters, calculate the depth of laundry hamper (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Type in exactly one number as your reply.
1.4
53
single_view
fill
depth_prediction_oc
scannet
Knowing blinds (red point) is at 3.0 meters, calculate the depth of monitor (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
1.4
54
single_view
fill
depth_prediction_oc
scannet
If microwave (red point) has a depth of 2.0 meters, how deep is the kitchen counter (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one numeric value is required for your response.
2.0
55
single_view
fill
depth_prediction_oc
scannet
Using sink (red point) at 1.5 meters as a guide, determine the depth of trash bin (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Please give a response in the form of a single number.
1.4
56
single_view
fill
depth_prediction_oc
scannet
The pillow (red point) lies at a depth of 2.0 meters. How deep is the center of lamp (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Ensure your answer contains only one number.
1.8
57
single_view
fill
depth_prediction_oc
scannet
At a depth of 2.3 meters, the chair (red point) serves as a reference. What is the depth of chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
1.4
58
single_view
fill
depth_prediction_oc
scannet
The camera measures computer tower (red point) at 1.9 meters depth. Predict the depth of the chair (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.4
59
single_view
fill
depth_prediction_oc
scannet
The chair (red point) lies at a depth of 2.4 meters. How deep is the center of backpack (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide a numeric response with just one value.
3.9
60
single_view
fill
depth_prediction_oc
scannet
With the table (red point) at 1.3 meters, determine the depth of chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Please respond with a single number only.
1.6
61
single_view
fill
depth_prediction_oc
scannet
If the armchair (red point) is positioned at 2.0 meters, estimate the depth of the armchair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
2.3
62
single_view
fill
depth_prediction_oc
scannet
With bar (red point) at 1.7 meters, calculate the depth of toilet (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
1.8
63
single_view
fill
depth_prediction_oc
scannet
Given backpack (red point) is positioned at 1.0 meters, how deep is the chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Your response should consist of just one numeric value.
2.4
64
single_view
fill
depth_prediction_oc
scannet
At 1.5 meters, the backpack (red point) sets the depth reference. Calculate the depth of the chair (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
2.6
65
single_view
fill
depth_prediction_oc
scannet
With the telephone (red point) having a depth of 0.8 meters, determine the depth of toilet paper (blue point) in the same frame. Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter one numeric value as your answer.
1.0
66
single_view
fill
depth_prediction_oc
scannet
At 3.2 meters, the ottoman (red point) sets the depth reference. Calculate the depth of the pillow (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
2.9
67
single_view
fill
depth_prediction_oc
scannet
The chair (red point) at 2.1 meters serves as a depth reference. What is the approximate depth of the chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Type in exactly one number as your reply.
1.9
68
single_view
fill
depth_prediction_oc
scannet
With the stool (red point) center located at 4.9 meters in the camera Z-axis(camera coordinate system), determine the depth of ping pong table (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
2.4
69
single_view
fill
depth_prediction_oc
scannet
Using bag (red point) as a reference with its center at 2.4 meters depth, calculate the depth of case of water bottles (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide a numeric response with just one value.
1.5
70
single_view
fill
depth_prediction_oc
scannet
The camera coordinate system positions bag (red point) at a depth of 2.0 meters. Predict the depth of tube (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
1.1
71
single_view
fill
depth_prediction_oc
scannet
Knowing case of water bottles (red point) is at 1.6 meters, calculate the depth of bin (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
2.7
72
single_view
fill
depth_prediction_oc
scannet
The depth of display case (red point) is 4.1 meters. Based on this, what is the approximate depth of table (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter one numeric value as your answer.
2.9
73
single_view
fill
depth_prediction_oc
scannet
The camera measures monitor (red point) at 1.9 meters depth. Predict the depth of the monitor (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Your response should consist of just one numeric value.
2.7
74
single_view
fill
depth_prediction_oc
scannet
The headphones (red point) is positioned at 1.6 meters depth. Predict the depth of the book (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
2.2
75
single_view
fill
depth_prediction_oc
scannet
Using monitor (red point) at 3.4 meters as a guide, determine the depth of keyboard (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
2.7
76
single_view
fill
depth_prediction_oc
scannet
At a depth of 2.5 meters, the seating (red point) serves as a reference. What is the depth of chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
1.7
77
single_view
fill
depth_prediction_oc
scannet
If the whiteboard (red point) lies at a depth of 3.1 meters, how deep is the center of the backpack (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Input a single number to complete your answer.
2.1
78
single_view
fill
depth_prediction_oc
scannet
With the window (red point) center located at 2.7 meters in the camera Z-axis(camera coordinate system), determine the depth of clothes (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your response as one numeric value only.
2.3
79
single_view
fill
depth_prediction_oc
scannet
The table (red point) at a depth of 2.4 meters serves as a reference. How deep is the chair (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Your input must be limited to a single number.
3.1
80
single_view
fill
depth_prediction_oc
scannet
With book (red point) at a depth of 1.7 meters, determine the approximate depth of book (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Your input must be limited to a single number.
1.1
81
single_view
fill
depth_prediction_oc
scannet
If shoes (red point) has a depth of 2.4 meters, how deep is the towel (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Please give a response in the form of a single number.
2.1
82
single_view
fill
depth_prediction_oc
scannet
The cooking pan (red point) lies at 3.8 meters depth. Predict the depth of the kitchen cabinets (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. The answer should be a single number, nothing more.
2.7
83
single_view
fill
depth_prediction_oc
scannet
If the dish rack (red point) lies at a depth of 2.0 meters, how deep is the center of the kitchen cabinets (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one numeric value is required for your response.
1.6
84
single_view
fill
depth_prediction_oc
scannet
Given the telephone (red point) is located at a depth of 2.5 meters in the Z-axis of the camera coordinate system, how far in depth is the chair (blue point) at its center? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
3.3
85
single_view
fill
depth_prediction_oc
scannet
If the center of the chair (red point) is 1.3 meters deep, what is the depth of box (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Ensure your answer contains only one number.
2.6
86
single_view
fill
depth_prediction_oc
scannet
The center of the bottle (red point) is positioned at 1.1 meters. Based on this, predict the depth of the center of the container (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.3
87
single_view
fill
depth_prediction_oc
scannet
The depth of the chair (red point) at its center is measured as 3.1 meters. Estimate the depth of the chair (blue point) at its center. Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
5.1
88
single_view
fill
depth_prediction_oc
scannet
The box (red point) at a depth of 2.5 meters serves as a reference. How deep is the copier (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.5
89
single_view
fill
depth_prediction_oc
scannet
The keyboard (red point) is placed at a depth of 1.1 meters along the Z-axis(camera coordinate system). Estimate the depth of monitor (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Answer using a single number and nothing else.
1.1
90
single_view
fill
depth_prediction_oc
scannet
The chair (red point) is 2.7 meters deep in the camera coordinate system. Predict the depth of the chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.9
91
single_view
fill
depth_prediction_oc
scannet
The refrigerator (red point) is placed at a depth of 3.7 meters along the Z-axis(camera coordinate system). Estimate the depth of vacuum cleaner (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide your input as exactly one number.
2.9
92
single_view
fill
depth_prediction_oc
scannet
Knowing that shelf (red point) is located at 2.9 meters, what is the depth of shelf (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide a numeric response with just one value.
2.8
93
single_view
fill
depth_prediction_oc
scannet
Given shelf (red point) is positioned at 1.8 meters, how deep is the compost bin (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.9
94
single_view
fill
depth_prediction_oc
scannet
The chair (red point) is 2.4 meters deep in the camera coordinate system. Predict the depth of the chair (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
1.1
95
single_view
fill
depth_prediction_oc
scannet
The towel (red point) is situated at 2.1 meters in the camera's Z-axis(camera coordinate system). What is the depth of the shelf (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one numeric value is required for your response.
2.3
96
single_view
fill
depth_prediction_oc
scannet
Given kitchen cabinet (red point) is positioned at 3.2 meters, how deep is the kitchen island (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Enter a single numeric response and no other details.
1.9
97
single_view
fill
depth_prediction_oc
scannet
Given that picture (red point) is at 2.8 meters, predict the depth of table (blue point). Calculate or judge based on the 3D center points of these objects. The unit is meter. Provide only one number for your response.
3.3
98
single_view
fill
depth_prediction_oc
scannet
The vending machine (red point) is placed at a depth of 4.6 meters along the Z-axis(camera coordinate system). Estimate the depth of laundry hamper (blue point) and Calculate or judge based on the 3D center points of these objects. The unit is meter. Only one number should be included in your response.
2.4
99
single_view
fill
depth_prediction_oc
scannet
The soap dish (red point) lies at a depth of 1.1 meters. How deep is the center of ledge (blue point)? Calculate or judge based on the 3D center points of these objects. The unit is meter. Submit your answer as one number only.
1.1
End of preview. Expand in Data Studio

GitHub Code arXiv Website

🎯 Spatial Perception And Reasoning Benchmark (SPAR-Bench)

A benchmark to evaluate spatial perception and reasoning in vision-language models (VLMs), with high-quality QA across 20 diverse tasks.

SPAR-Bench is a high-quality benchmark for evaluating spatial perception and reasoning in vision-language models (VLMs). It covers 20 diverse spatial tasks across single-view, multi-view, and video settings, with a total of 7,207 manually verified QA pairs.

SPAR-Bench is derived from the large-scale SPAR-7M dataset, and specifically designed to support zero-shot evaluation and task-specific analysis

📌 SPAR-Bench at a glance:

  • ✅ 7,207 manually verified QA pairs
  • 🧠 20 spatial tasks (depth, distance, relation, imagination, etc.)
  • 🎥 Supports single-view, multi-view inputs
  • 📏 Two evaluation metrics: Accuracy & MRA
  • 📷 Available in RGB-only and RGB-D versions

🧱 Available Variants

We provide four versions of SPAR-Bench, covering both RGB-only and RGB-D settings, as well as full-size and lightweight variants:

Dataset Name Description
SPAR-Bench Full benchmark (7,207 QA) with RGB images
SPAR-Bench-RGBD Full benchmark with depths, camera pose and intrinsics
SPAR-Bench-Tiny 1,000-sample subset (50 QA per task), for fast evaluation or APIs
SPAR-Bench-Tiny-RGBD Tiny version with RGBD inputs

🔎 Tiny versions are designed for quick evaluation (e.g., APIs, human studies).
💡 RGBD versions include depths, poses, and intrinsics, suitable for 3D-aware models.

To load a different version via datasets, simply change the dataset name:

from datasets import load_dataset
spar = load_dataset("jasonzhango/SPAR-Bench")
spar_rgbd = load_dataset("jasonzhango/SPAR-Bench-RGBD")
spar_tiny = load_dataset("jasonzhango/SPAR-Bench-Tiny")
spar_tiny_rgbd = load_dataset("jasonzhango/SPAR-Bench-Tiny-RGBD")

🕹️ Evaluation

SPAR-Bench supports two evaluation metrics, depending on the question type:

  • Accuracy – for multiple-choice questions (exact match)
  • Mean Relative Accuracy (MRA) – for numerical-answer questions (e.g., depth, distance)

🧠 The MRA metric is inspired by the design in Thinking in Space, and is tailored for spatial reasoning tasks involving quantities like distance and depth.

We provide an evaluation pipeline in our GitHub repository, built on top of lmms-eval.

📚 Bibtex

If you find this project or dataset helpful, please consider citing our paper:

@article{zhang2025from,
    title={From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D},
    author={Zhang, Jiahui and Chen, Yurui and Zhou, Yanpeng and Xu, Yueming and Huang, Ze and Mei, Jilin and Chen, Junhui and Yuan, Yujie and Cai, Xinyue and Huang, Guowei and Quan, Xingyue and Xu, Hang and Zhang, Li},
    year={2025},
    journal={arXiv preprint arXiv:2503.22976},
}
Downloads last month
190