id
int64 968
4.38k
| dataset
stringclasses 3
values | scene_name
stringlengths 8
12
| question_type
stringclasses 3
values | category
null | question
stringlengths 812
833
| options
sequencelengths 2
4
| ground_truth
stringclasses 7
values | mc_answer
stringclasses 3
values | video_suffix
stringclasses 2
values | version
stringclasses 1
value | category, first object
null | category, second object
null | category, positioning object
stringlengths 2
14
| category, orienting object
stringlengths 2
14
| category, querying object
stringlengths 2
14
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
968 | arkitscenes | 42444976 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'table', 'dishwasher'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. front-right",
"C. back-left",
"D. back-right"
] | front-left | A | .mov | 241009 | null | null | refrigerator | table | dishwasher |
987 | arkitscenes | 42446167 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['table', 'bed', 'toilet'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-left",
"B. back-right",
"C. front-left",
"D. front-right"
] | back-left | A | .mov | 241009 | null | null | table | bed | toilet |
1,023 | arkitscenes | 45261121 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['table', 'stool', 'tv'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-right",
"B. front-right",
"C. back-left",
"D. front-left"
] | front-right | B | .mov | 241009 | null | null | table | stool | tv |
1,036 | arkitscenes | 47204552 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['stove', 'tv', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-left",
"B. front-right",
"C. back-right",
"D. front-left"
] | front-right | B | .mov | 241009 | null | null | stove | tv | table |
1,049 | arkitscenes | 47331970 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'table', 'dishwasher'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-right",
"B. back-left",
"C. front-left",
"D. front-right"
] | front-left | C | .mov | 241009 | null | null | refrigerator | table | dishwasher |
1,059 | arkitscenes | 47333899 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'table', 'washer'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-right",
"B. front-right",
"C. back-left",
"D. front-left"
] | back-left | C | .mov | 241009 | null | null | refrigerator | table | washer |
1,125 | arkitscenes | 42445031 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['tv', 'bed', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left",
"C. back"
] | right | A | .mov | 241009 | null | null | tv | bed | table |
1,149 | arkitscenes | 42898527 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['stove', 'dishwasher', 'refrigerator'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. back",
"C. left"
] | right | A | .mov | 241009 | null | null | stove | dishwasher | refrigerator |
1,164 | arkitscenes | 45260928 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['tv', 'sofa', 'fireplace'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left",
"C. back"
] | left | B | .mov | 241009 | null | null | tv | sofa | fireplace |
1,199 | arkitscenes | 47333899 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'stove', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. back",
"C. right"
] | left | A | .mov | 241009 | null | null | refrigerator | stove | table |
1,213 | arkitscenes | 47334239 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['bathtub', 'washer', 'stool'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. right",
"C. left"
] | left | C | .mov | 241009 | null | null | bathtub | washer | stool |
1,215 | arkitscenes | 47334239 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['washer', 'bathtub', 'stool'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | right | C | .mov | 241009 | null | null | washer | bathtub | stool |
1,235 | arkitscenes | 47430468 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['stove', 'stool', 'refrigerator'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. right",
"C. left"
] | right | B | .mov | 241009 | null | null | stove | stool | refrigerator |
1,271 | arkitscenes | 42897688 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['bed', 'stool', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | right | B | .mov | 241009 | null | null | bed | stool | table |
1,292 | arkitscenes | 47331970 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['dishwasher', 'refrigerator', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | left | A | .mov | 241009 | null | null | dishwasher | refrigerator | table |
1,321 | arkitscenes | 47429922 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['sofa', 'tv', 'chair'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | right | B | .mov | 241009 | null | null | sofa | tv | chair |
1,519 | scannetpp | acd95847c5 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['door', 'bookshelf', 'laptop'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right",
"C. back"
] | left | A | .mp4 | 241009 | null | null | door | bookshelf | laptop |
1,528 | scannetpp | a24f64f7fb | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['monitor', 'computer mouse', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | right | C | .mp4 | 241009 | null | null | monitor | computer mouse | door |
1,531 | scannetpp | 5eb31827b7 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['computer tower', 'telephone', 'keyboard'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. back",
"C. left"
] | left | C | .mp4 | 241009 | null | null | computer tower | telephone | keyboard |
1,554 | scannetpp | 3864514494 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['heater', 'door', 'microwave'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. right",
"C. left"
] | right | B | .mp4 | 241009 | null | null | heater | door | microwave |
1,560 | scannetpp | e398684d27 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['door', 'ceiling light', 'monitor'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. right",
"C. left"
] | left | C | .mp4 | 241009 | null | null | door | ceiling light | monitor |
1,570 | scannetpp | 286b55a2bf | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['toilet', 'exhaust fan', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | left | B | .mp4 | 241009 | null | null | toilet | exhaust fan | door |
1,571 | scannetpp | a980334473 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['ceiling light', 'chair', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. back",
"C. left"
] | right | A | .mp4 | 241009 | null | null | ceiling light | chair | door |
1,575 | scannetpp | f2dc06b1d2 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['chair', 'table', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. back",
"C. right"
] | right | C | .mp4 | 241009 | null | null | chair | table | door |
1,582 | scannetpp | bcd2436daf | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['chair', 'pillow', 'ceiling light'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | right | C | .mp4 | 241009 | null | null | chair | pillow | ceiling light |
1,584 | scannetpp | 0d2ee665be | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['door', 'cutting board', 'computer mouse'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right",
"C. back"
] | right | B | .mp4 | 241009 | null | null | door | cutting board | computer mouse |
1,838 | scannetpp | 5eb31827b7 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['door', 'telephone', 'keyboard'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-left",
"B. front-left",
"C. back-right",
"D. front-right"
] | front-left | B | .mp4 | 241009 | null | null | door | telephone | keyboard |
1,855 | scannetpp | 40aec5fffa | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'heater', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. front-right",
"C. back-right",
"D. back-left"
] | back-right | C | .mp4 | 241009 | null | null | refrigerator | heater | door |
2,440 | scannetpp | 3e8bba0176 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['coat hanger', 'paper bag', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left"
] | left | B | .mp4 | 241009 | null | null | coat hanger | paper bag | door |
2,478 | scannetpp | 09c1414f1b | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['shoe rack', 'microwave', 'kettle'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left"
] | right | A | .mp4 | 241009 | null | null | shoe rack | microwave | kettle |
3,585 | scannet | scene0378_01 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['window', 'clock', 'trash bin'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. front-right",
"C. back-left",
"D. back-right"
] | front-right | B | .mp4 | 241009 | null | null | window | clock | trash bin |
3,586 | scannet | scene0378_01 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['window', 'trash bin', 'clock'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. back-left",
"C. back-right",
"D. front-right"
] | front-left | A | .mp4 | 241009 | null | null | window | trash bin | clock |
3,657 | scannet | scene0608_00 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['sofa', 'lamp', 'guitar'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. back-right",
"C. front-right",
"D. back-left"
] | front-left | A | .mp4 | 241009 | null | null | sofa | lamp | guitar |
3,666 | scannet | scene0050_01 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['printer', 'lamp', 'piano'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back-right",
"B. front-left",
"C. back-left",
"D. front-right"
] | front-left | B | .mp4 | 241009 | null | null | printer | lamp | piano |
3,697 | scannet | scene0246_00 | object_rel_direction_v1 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['nightstand', 'chair', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. front-left",
"B. front-right",
"C. back-right",
"D. back-left"
] | front-right | B | .mp4 | 241009 | null | null | nightstand | chair | table |
3,711 | scannet | scene0357_01 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['window', 'table', 'towel'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | left | A | .mp4 | 241009 | null | null | window | table | towel |
3,712 | scannet | scene0357_01 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['towel', 'window', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | left | A | .mp4 | 241009 | null | null | towel | window | table |
3,716 | scannet | scene0629_01 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['bed', 'chair', 'mirror'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left"
] | left | B | .mp4 | 241009 | null | null | bed | chair | mirror |
3,728 | scannet | scene0699_00 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['chair', 'nightstand', 'closet'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. right"
] | right | B | .mp4 | 241009 | null | null | chair | nightstand | closet |
3,730 | scannet | scene0699_00 | object_rel_direction_v3 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['nightstand', 'chair', 'closet'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left"
] | left | B | .mp4 | 241009 | null | null | nightstand | chair | closet |
4,240 | scannet | scene0304_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['backpack', 'trash bin', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. right",
"C. left"
] | back | A | .mp4 | 241009 | null | null | backpack | trash bin | door |
4,243 | scannet | scene0307_02 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['refrigerator', 'chair', 'table'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | left | B | .mp4 | 241009 | null | null | refrigerator | chair | table |
4,298 | scannet | scene0697_01 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['mirror', 'closet', 'trash bin'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. left",
"C. back"
] | left | B | .mp4 | 241009 | null | null | mirror | closet | trash bin |
4,316 | scannet | scene0458_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['window', 'door', 'towel'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | left | B | .mp4 | 241009 | null | null | window | door | towel |
4,330 | scannet | scene0593_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['backpack', 'fan', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. back",
"C. left"
] | right | A | .mp4 | 241009 | null | null | backpack | fan | door |
4,359 | scannet | scene0144_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['lamp', 'printer', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. right",
"B. back",
"C. left"
] | left | C | .mp4 | 241009 | null | null | lamp | printer | door |
4,362 | scannet | scene0144_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['window', 'lamp', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. back",
"B. left",
"C. right"
] | right | C | .mp4 | 241009 | null | null | window | lamp | door |
4,364 | scannet | scene0144_00 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['lamp', 'window', 'door'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. back",
"C. right"
] | left | A | .mp4 | 241009 | null | null | lamp | window | door |
4,376 | scannet | scene0221_01 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['microwave', 'nightstand', 'trash bin'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. back",
"C. right"
] | right | C | .mp4 | 241009 | null | null | microwave | nightstand | trash bin |
4,381 | scannet | scene0222_01 | object_rel_direction_v2 | null | [Task]
This video captures an indoor scene. Your task is to first understand the spatial layout of the scene and then represent it using a 10x10 matrix.
[Rule]
1. We provide the cared categories and their corresponding instances: ['lamp', 'radiator', 'nightstand'] in this scene. Focus ONLY on these categories.
2. You should use 0 to represent an empty grid and different numbers to represent different instances.
3. If a category contains multiple instances, include all of them in the matrix.
4. If objects overlap in the same grid, retain only the object with the largest area.
[Output]
Fill in the locations of ALL objects (at the instance level) in the matrix.
STRICTLY follow this exact JSON format for your response:
{"grid": 10x10 matrix (list of lists), "legend": {0: "empty", 1: "object 1", 2: "object 2", ...}} | [
"A. left",
"B. back",
"C. right"
] | back | B | .mp4 | 241009 | null | null | lamp | radiator | nightstand |