task
stringclasses 1
value | source
stringclasses 1
value | video
stringlengths 37
38
| duration
int64 120
150
| tgt
listlengths 1
24
| conversations
listlengths 2
2
|
|---|---|---|---|---|---|
tvg
|
qvhighlights
|
qvh/videos/5LPS7E-0tuA_60.0_210.0.mp4
| 150
|
[
[
86,
100
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A helicopter is flying over the area against a clear blue sky'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/M9Qcylr7sXY_60.0_210.0.mp4
| 150
|
[
[
80,
100
],
[
108,
112
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A happy couple both wearing gray sitting in a car together smiling and talking' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/M9Qcylr7sXY_210.0_360.0.mp4
| 150
|
[
[
28,
34
],
[
38,
56
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A couple walking on a beach and talking' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/M9Qcylr7sXY_360.0_510.0.mp4
| 150
|
[
[
6,
30
],
[
92,
106
],
[
116,
140
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Man and woman have conversations outside in a patio'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/g01YnqH-2ek_210.0_360.0.mp4
| 150
|
[
[
102,
104
],
[
108,
126
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Man in blazer and woman in striped shirt walk down the street together' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/g01YnqH-2ek_60.0_210.0.mp4
| 150
|
[
[
76,
86
],
[
88,
94
],
[
108,
110
],
[
122,
128
],
[
136,
142
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Four Asian women sitting at a table with champagne glasses and talking about fashion' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/JdrErPEIsr0_210.0_360.0.mp4
| 150
|
[
[
34,
44
],
[
48,
50
],
[
64,
70
],
[
78,
104
],
[
118,
122
],
[
130,
134
],
[
138,
146
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Bearded man talks to camera from his office' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/JdrErPEIsr0_360.0_510.0.mp4
| 150
|
[
[
130,
150
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Protestors are gathered in a smoky street holding up signs and flares' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/JdrErPEIsr0_60.0_210.0.mp4
| 150
|
[
[
10,
24
],
[
30,
58
],
[
114,
136
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Woman stands perfectly between two trees while talking', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/UWjgV9nYBi4_210.0_360.0.mp4
| 150
|
[
[
128,
150
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A woman in a hood and mask is browsing the snack aisle of a grocery store', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/UWjgV9nYBi4_360.0_510.0.mp4
| 150
|
[
[
2,
6
],
[
8,
26
],
[
28,
32
],
[
44,
46
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Muslim woman wears mask while grocery shopping' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/UWjgV9nYBi4_60.0_210.0.mp4
| 150
|
[
[
94,
110
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A lite up arcade wheel spins around'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/Wgp6Yzgn7nY_60.0_210.0.mp4
| 150
|
[
[
0,
150
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Old man speaks into a professional mic', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/Wgp6Yzgn7nY_360.0_510.0.mp4
| 144
|
[
[
0,
144
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'A man talks into a mic sat at home with some pictures on the wall behind him' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/Wgp6Yzgn7nY_210.0_360.0.mp4
| 150
|
[
[
0,
150
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Old man speaks into a professional mic', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/2b9txcAt4e0_60.0_210.0.mp4
| 150
|
[
[
112,
138
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A crowd of people is walking in mud' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/2b9txcAt4e0_360.0_510.0.mp4
| 150
|
[
[
0,
24
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Asian women wear traditional garb while holding onto yellow umbrellas' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/2b9txcAt4e0_210.0_360.0.mp4
| 150
|
[
[
122,
146
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'There are large mountains in the distance' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/XFrOEO9GUF8_60.0_210.0.mp4
| 150
|
[
[
98,
112
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Two women sit really close on the bench together' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/O8v5zRFHGh0_360.0_510.0.mp4
| 150
|
[
[
50,
86
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Little girl ready to be a youtuber' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/O8v5zRFHGh0_60.0_210.0.mp4
| 150
|
[
[
24,
36
],
[
38,
42
],
[
46,
48
],
[
58,
68
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Woman vlogs lying down on the bed' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/O8v5zRFHGh0_210.0_360.0.mp4
| 150
|
[
[
34,
68
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Woman in gray monologues from her bed'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/NRjvpWJxuaw_60.0_210.0.mp4
| 150
|
[
[
22,
38
],
[
40,
46
],
[
48,
52
],
[
54,
66
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'a woman with an indian accent is showing her hotel room' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/NRjvpWJxuaw_210.0_360.0.mp4
| 150
|
[
[
120,
150
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'While on a train, a vlogger talks about her coming Nagaland series and Namibia series of videos' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/KDkmbZNEo4I_360.0_510.0.mp4
| 150
|
[
[
0,
4
],
[
52,
56
],
[
90,
94
],
[
114,
118
],
[
132,
134
],
[
138,
144
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'two girls both hoop earrings showing off their outfit and head to dinner with a group of friends in new york city, time square', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/KDkmbZNEo4I_60.0_210.0.mp4
| 150
|
[
[
128,
150
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Beautiful young American tourist enjoying on the street of new York' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/KDkmbZNEo4I_210.0_360.0.mp4
| 150
|
[
[
0,
30
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Woman in red top walks down the street' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/x-l6vBbHpVk_60.0_210.0.mp4
| 150
|
[
[
64,
122
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Man is standing around talking wearing a bike helmet' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/x-l6vBbHpVk_360.0_510.0.mp4
| 142
|
[
[
0,
8
],
[
12,
20
],
[
28,
34
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'a man and showing is showing in the pool swimming, then are back in the hotel room with a shirtless man talking' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/x-l6vBbHpVk_210.0_360.0.mp4
| 150
|
[
[
16,
34
],
[
82,
116
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'A man in black tank top is talking in front a camera' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/EbpURkWDQLM_60.0_210.0.mp4
| 150
|
[
[
48,
64
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Julian Assange hides out at the Ecuadorian Embassy in London', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/EbpURkWDQLM_360.0_510.0.mp4
| 150
|
[
[
6,
24
],
[
44,
60
],
[
74,
78
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Graphics show the money raised in political campaigns' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/EbpURkWDQLM_210.0_360.0.mp4
| 150
|
[
[
6,
22
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Different headlines are about Turkey and Saudi Arabia'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/llNpCfAjhj4_60.0_210.0.mp4
| 150
|
[
[
74,
102
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Man in yellow shirt is driving a car'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/llNpCfAjhj4_210.0_360.0.mp4
| 150
|
[
[
88,
92
],
[
98,
108
],
[
114,
122
],
[
138,
150
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A travel blogger's selfie video with ocean water in the background' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/llNpCfAjhj4_360.0_510.0.mp4
| 150
|
[
[
68,
122
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A man with no shirt holds a snake in his hands'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/llNpCfAjhj4_510.0_660.0.mp4
| 150
|
[
[
70,
110
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'a woman helps remove cactus spines from a mans leg' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nCttqSbHIig_210.0_360.0.mp4
| 150
|
[
[
0,
20
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Vlogger holds coins in their hands' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nCttqSbHIig_360.0_510.0.mp4
| 150
|
[
[
18,
30
],
[
40,
58
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A person checking a coin exchange rate card' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nCttqSbHIig_60.0_210.0.mp4
| 150
|
[
[
18,
44
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Cute little girl sharing her breakfast menu' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nCttqSbHIig_510.0_660.0.mp4
| 150
|
[
[
42,
88
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Many children are scrambling for candy during an Easter egg hunt' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nCttqSbHIig_660.0_810.0.mp4
| 150
|
[
[
18,
52
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'A family shops for a grill at walmart' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/sP_RRzYdCZ0_60.0_210.0.mp4
| 150
|
[
[
4,
12
],
[
40,
48
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'People yell at the police from behind a white fence' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/sP_RRzYdCZ0_360.0_510.0.mp4
| 150
|
[
[
70,
76
],
[
88,
102
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'A soldier is wearing a beret during his interview' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/sP_RRzYdCZ0_210.0_360.0.mp4
| 150
|
[
[
70,
100
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A woman in orange blouse is talking to a man in light blue shirt in front of the camera' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/gpGDn-CAG_Y_60.0_210.0.mp4
| 150
|
[
[
86,
106
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'The pope is making his way down red stairs from a plane'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/gpGDn-CAG_Y_210.0_360.0.mp4
| 138
|
[
[
92,
122
],
[
128,
138
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'The pope is sitting on a fancy chair giving an inteview', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3h8XL2brYHg_210.0_360.0.mp4
| 150
|
[
[
2,
4
],
[
22,
26
],
[
46,
54
],
[
62,
64
],
[
76,
84
],
[
116,
124
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Different still photos are shown of Donald Trump' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3h8XL2brYHg_60.0_210.0.mp4
| 150
|
[
[
118,
126
],
[
130,
142
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A 177,000 year old fossilized human jaw bone is displayed', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3h8XL2brYHg_360.0_510.0.mp4
| 150
|
[
[
0,
28
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Donald Trump holds a bipartisan meeting' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/QmvKWmTAwvU_60.0_210.0.mp4
| 150
|
[
[
42,
56
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Woman with blue jacket stands in the middle of the street' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/QmvKWmTAwvU_210.0_360.0.mp4
| 150
|
[
[
8,
10
],
[
14,
26
],
[
36,
46
],
[
64,
66
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'People are traversing the wilderness', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/q-OF2n0tg90_210.0_360.0.mp4
| 150
|
[
[
0,
36
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Man wears striped shirt sitting on the bed', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/q-OF2n0tg90_360.0_510.0.mp4
| 150
|
[
[
90,
138
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Woman and man vlog with a plate of food between them' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/q-OF2n0tg90_60.0_210.0.mp4
| 150
|
[
[
66,
70
],
[
72,
98
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A man is getting his hair cut at a barber shop' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/1NTwXCd6Qqk_60.0_210.0.mp4
| 150
|
[
[
32,
68
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A man is realeased from prison after decades inside and speaks to the public' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/1NTwXCd6Qqk_360.0_510.0.mp4
| 150
|
[
[
28,
38
],
[
40,
44
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Donald Trump speaks in a ceremony at home' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/1NTwXCd6Qqk_210.0_360.0.mp4
| 150
|
[
[
8,
24
],
[
34,
38
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Paul Ryan speaks at a podium surrounded by American flags', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/yRyDYz33EU0_60.0_210.0.mp4
| 150
|
[
[
66,
78
],
[
86,
94
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'View as seen from the airplane window' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/yRyDYz33EU0_360.0_510.0.mp4
| 150
|
[
[
100,
104
],
[
106,
116
],
[
118,
144
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Teen girls are hanging out in a tropical restaurant' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/yRyDYz33EU0_210.0_360.0.mp4
| 150
|
[
[
112,
134
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Woman with lei holds up her camera in the bathroom' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/yRyDYz33EU0_660.0_810.0.mp4
| 150
|
[
[
44,
66
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A group of women in white robes are having a tour of the spa with massage tables' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/yRyDYz33EU0_510.0_660.0.mp4
| 150
|
[
[
88,
114
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'a woman talks while seated in front of vanity lighting' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/_XZn6wIojo0_210.0_360.0.mp4
| 150
|
[
[
66,
82
],
[
84,
90
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Man holds a piece of bread and eats it' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/_XZn6wIojo0_360.0_510.0.mp4
| 150
|
[
[
0,
8
],
[
10,
82
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'American tourist couple showing some really beautiful scene from Romania' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/_XZn6wIojo0_60.0_210.0.mp4
| 150
|
[
[
30,
62
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A video capturing a hill and water'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/8oGkr6EyZ1Q_210.0_360.0.mp4
| 150
|
[
[
84,
138
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Young people walk through the city at night'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/8oGkr6EyZ1Q_360.0_510.0.mp4
| 150
|
[
[
34,
70
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Little girl making drawing with her father and making jokes' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/8oGkr6EyZ1Q_60.0_210.0.mp4
| 150
|
[
[
34,
44
],
[
46,
48
],
[
56,
68
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A baby is a carrier, then moved to a stroller and then a car seat'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3LG_pR4R5KY_210.0_360.0.mp4
| 150
|
[
[
46,
74
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A man is playing and laughing with its cute baby', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3LG_pR4R5KY_360.0_510.0.mp4
| 150
|
[
[
130,
144
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A woman gets sandals out of the bag' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3LG_pR4R5KY_60.0_210.0.mp4
| 150
|
[
[
76,
86
],
[
90,
116
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Boy crawls around on the couch' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3LG_pR4R5KY_510.0_660.0.mp4
| 150
|
[
[
72,
94
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'A woman is helping a baby play with a toy on the living room floor' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3LG_pR4R5KY_660.0_810.0.mp4
| 150
|
[
[
108,
140
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A woman is looking through various food containers in the kitchen and a gingerbread house', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/bnBAKvYSLBM_210.0_360.0.mp4
| 150
|
[
[
96,
128
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A woman is testing different make up colors on her hands and showing them in the bottles' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/bnBAKvYSLBM_360.0_510.0.mp4
| 150
|
[
[
106,
126
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Man in white top cuts up a watermelon'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/bnBAKvYSLBM_60.0_210.0.mp4
| 150
|
[
[
4,
32
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Cart worker makes vlogger some food' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3tplGfPWVg4_360.0_510.0.mp4
| 150
|
[
[
128,
150
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'Overhead shot of dinner on a white plate' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3tplGfPWVg4_60.0_210.0.mp4
| 150
|
[
[
0,
20
],
[
32,
64
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A woman with black long hair is walking towards a bank', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/3tplGfPWVg4_210.0_360.0.mp4
| 150
|
[
[
40,
104
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'People pick out different foods from a buffet and show their selections at the table while eating' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/q7tastD4eYA_60.0_210.0.mp4
| 150
|
[
[
0,
150
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'The host of the show with a mic in front of him is talking in front of the camera' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/4rACNtuimq4_60.0_210.0.mp4
| 150
|
[
[
0,
36
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Woman in green monologues in her car', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/4rACNtuimq4_360.0_510.0.mp4
| 150
|
[
[
4,
40
],
[
70,
122
]
] |
[
{
"from": "human",
"value": "<image>\nLocate the visual content mentioned in the text query 'A woman films while giving a tour of her cruise ship cabin room' within the video using timestamps.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/4rACNtuimq4_210.0_360.0.mp4
| 150
|
[
[
72,
116
],
[
128,
144
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'Woman in yellow lop lounges on a blue deck chair'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/mcb_rWj0fYA_60.0_210.0.mp4
| 150
|
[
[
60,
90
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A young woman with a ponytail describing a room', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/mcb_rWj0fYA_60.0_210.0.mp4
| 150
|
[
[
96,
150
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'A girl showing the organizing of her dorm room' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/mcb_rWj0fYA_210.0_360.0.mp4
| 150
|
[
[
6,
32
]
] |
[
{
"from": "human",
"value": "<image>\nDetect and report the start and end timestamps of the video segment that semantically matches the given textual query 'A girl organising her dresses in a room'.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/mcb_rWj0fYA_210.0_360.0.mp4
| 150
|
[
[
0,
28
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'A young girl are organizing a room and hanging some clothing up' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/-SwWStoWYKo_360.0_510.0.mp4
| 150
|
[
[
128,
142
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A racoon is sitting in a chair', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/-SwWStoWYKo_60.0_210.0.mp4
| 150
|
[
[
120,
150
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'A girl captures stunning view of a mall which resembles a library shelf and the food in the mall' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/-SwWStoWYKo_210.0_360.0.mp4
| 150
|
[
[
48,
72
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Asian woman eats with chopsticks' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/USGN-aTzlG4_60.0_210.0.mp4
| 150
|
[
[
82,
110
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A woman shows how she has packed her suitcases which are laying on the floor', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/USGN-aTzlG4_210.0_360.0.mp4
| 150
|
[
[
0,
26
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Cat is little depress to sea off her owner' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/USGN-aTzlG4_360.0_510.0.mp4
| 150
|
[
[
48,
78
],
[
82,
96
],
[
106,
138
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'Woman walks around a mall wearing a black vest' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/ytaP1WKArRs_60.0_210.0.mp4
| 150
|
[
[
72,
110
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'Girl walking with her dog', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/ytaP1WKArRs_360.0_510.0.mp4
| 150
|
[
[
26,
150
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'Girl sharing her makeup stuff on camera' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/ytaP1WKArRs_210.0_360.0.mp4
| 150
|
[
[
64,
82
],
[
84,
92
]
] |
[
{
"from": "human",
"value": "<image>\nThe given natural language query 'Woman holds her mask to the side when she is talking' is semantically aligned with a video moment, please give the start time and end time of the video moment.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nqpx0jIZb34_210.0_360.0.mp4
| 150
|
[
[
0,
68
],
[
70,
82
],
[
84,
150
]
] |
[
{
"from": "human",
"value": "<image>\nGiven a textual query: 'A woman in blue dotted blouse is talking while cooking', when does the described content occur in the video? Please return the timestamp in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nqpx0jIZb34_60.0_210.0.mp4
| 150
|
[
[
90,
120
]
] |
[
{
"from": "human",
"value": "<image>\nLocalize the visual content described by the given textual query 'A woman is preparing a pot of chickpeas' in the video, and output the start and end timestamps in seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
tvg
|
qvhighlights
|
qvh/videos/nqpx0jIZb34_360.0_510.0.mp4
| 150
|
[
[
0,
150
]
] |
[
{
"from": "human",
"value": "<image>\nFind the video segment that corresponds to the given textual query 'A video showing how to cook food' and determine its start and end seconds.The format of your response should be: 'The event happens in <start time> - <end time>''."
},
{
"from": "gpt",
"value": "The event happens in <vid> - <vid>."
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.