Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
ZihanWang314 commited on
Commit
4861b51
·
1 Parent(s): 46f4f70
Files changed (2) hide show
  1. README.md +88 -44
  2. assets/img/logo.png +0 -0
README.md CHANGED
@@ -60,50 +60,107 @@ configs:
60
  - config_name: default
61
  data_files:
62
  - split: train
63
- path: data/train-*
64
  - split: test
65
  path: data/test-*
66
  ---
67
 
68
  <h1 align='center' style="text-align:center; font-weight:bold; font-size:2.0em;letter-spacing:2.0px;">
69
- [[PAPER_TITLE]]</h1>
 
 
70
  <p align='center' style="text-align:center;font-size:1.25em;">
71
- <a href="[[AUTHOR_PAGE]]" target="_blank" style="text-decoration: none;">[[FIRST_NAME]]&nbsp;[[LAST_NAME]]</a>,&nbsp;
72
- &nbsp;Northwestern University<br/>
73
- <em>[[CONFERENCE_TITLE]]</em><br/>
74
- <a href="[[WEBSITE_LINK]]" title="Website" target="_blank" rel="nofollow" style="text-decoration: none;">🌎Website</a> |
75
- <a href="[[CODE_LINK]]" title="Dataset" target="_blank" rel="nofollow" style="text-decoration: none;">🧑‍💻Code</a> |
76
- <a href="[[ARXIV_LINK]]" title="aXiv" target="_blank" rel="nofollow" style="text-decoration: none;">📄arXiv</a> |
77
- <a href="[[LEADERBOARD_LINK]]" title="Leaderboard" target="_blank" rel="nofollow" style="text-decoration: none;">🏆 Leaderboard</a>
 
 
 
 
 
 
 
 
 
 
 
78
  </p>
79
 
80
-
81
- <h1 align='center' style="text-align:center; font-weight:bold; font-size:2.0em;letter-spacing:2.0px;">
82
- [[TITLE]]</h1>
83
  <p align='center' style="text-align:center;font-size:1.25em;">
84
- [[AUTHOR_1]],&nbsp;
85
- [[AUTHOR_2]],&nbsp;
86
- [[AUTHOR_3]],&nbsp;
87
- [[AUTHOR_4]],&nbsp;
88
- [[AUTHOR_5]],&nbsp;<br/>
89
- [[AUTHOR_6]],&nbsp;
90
- [[AUTHOR_7]],&nbsp;
91
- [[AUTHOR_8]],&nbsp;
92
- [[AUTHOR_9]],&nbsp;
93
- [[AUTHOR_10]]<br/>
94
- &nbsp;[[INSTITUTION]]<br/>
95
- <em>[[VENUE_AND_YEAR]]</em><br/>
96
- <a href="[[WEBSITE_LINK]]" title="Website" target="_blank" rel="nofollow" style="text-decoration: none;">🌎Website</a> |
97
- <a href="[[CODE_LINK]]" title="Dataset" target="_blank" rel="nofollow" style="text-decoration: none;">🧑‍💻Code</a> |
98
- <a href="[[PAPER_LINK]]" title="aXiv" target="_blank" rel="nofollow" style="text-decoration: none;">📄arXiv</a> |
99
- <a href="[[LEADERBOARD_LINK]]" title="Leaderboard" target="_blank" rel="nofollow" style="text-decoration: none;">🏆 Leaderboard</a>
100
  </p>
101
 
102
- ## 📣 News
103
 
104
- - **[[DATE]]: [[NEWS_ITEM]]**
105
 
106
- <img src="[[IMAGE_LINK]]" alt="image description" width="850" height="200">
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
107
 
108
  ## Abstract
109
 
@@ -113,20 +170,7 @@ configs:
113
 
114
  <img src="[[STATISTICS_IMAGE_LINK]]" alt="image description" width="850" height="200">
115
 
116
- ## Downloading Data and Benchmarking Guide
117
- To download the data:
118
-
119
- 1. **Download Instructions:**
120
- Follow the instructions provided [here]( [[DOWNLOAD_INSTRUCTIONS_LINK]] ).
121
- We provide unique identifiers (UIDs) for the data sets in ```[[DATA_ORGANIZATION_PATH]]```.
122
-
123
- 2. **Demo Data:**
124
- For a quick start, we include one demo example: ```[[DEMO_DATA_PATH]]```.
125
-
126
- 3. **Benchmarking Guide:**
127
- We provide a step-by-step guide for benchmarking [here]( [[BENCHMARKING_GUIDE_LINK]] ).
128
 
129
- **👉Note:** For evaluations, we recommend starting with our development set ([[DEV_SET_DETAILS]]).
130
 
131
  ## Dataset Organization
132
 
 
60
  - config_name: default
61
  data_files:
62
  - split: train
63
+ path: data/trai n-*
64
  - split: test
65
  path: data/test-*
66
  ---
67
 
68
  <h1 align='center' style="text-align:center; font-weight:bold; font-size:2.0em;letter-spacing:2.0px;">
69
+ LV-Haystack: Temporal Search in Long-Form Video Understanding</h1>
70
+
71
+
72
  <p align='center' style="text-align:center;font-size:1.25em;">
73
+ <a href="https://jhuiye.com/" target="_blank">Jinhui Ye<sup>1</sup></a>,&nbsp;
74
+ <a href="https://zihanwang314.github.io/" target="_blank">Zihan Wang<sup>2</sup></a>,&nbsp;
75
+ <a href="https://haosensun.github.io/" target="_blank">Haosen Sun<sup>2</sup></a>,&nbsp;
76
+ <a href="https://keshik6.github.io/" target="_blank">Keshigeyan Chandrasegaran<sup>1</sup></a>,&nbsp;
77
+ <a href="https://zanedurante.github.io/" target="_blank">Zane Durante<sup>1</sup></a>,&nbsp;<br/>
78
+ <a href="https://ceyzaguirre4.github.io/" target="_blank">Cristobal Eyzaguirre<sup>1</sup></a>,&nbsp;
79
+ <a href="https://talkingtorobots.com/yonatanbisk.html" target="_blank">Yonatan Bisk<sup>3</sup></a>,&nbsp;
80
+ <a href="https://www.niebles.net/" target="_blank">Juan Carlos Niebles<sup>1</sup></a>,&nbsp;
81
+ <a href="https://profiles.stanford.edu/ehsan-adeli" target="_blank">Ehsan Adeli<sup>1</sup></a>,&nbsp;
82
+ <a href="https://profiles.stanford.edu/fei-fei-li/" target="_blank">Li Fei-Fei<sup>1</sup></a>,&nbsp;
83
+ <a href="https://jiajunwu.com/" target="_blank">Jiajun Wu<sup>1</sup></a>,&nbsp;
84
+ <a href="https://limanling.github.io/" target="_blank">Manling Li<sup>2</sup></a><br/>
85
+ &nbsp;Stanford University<sup>1</sup>, Northwestern University<sup>2</sup>, Carnegie Mellon University<sup>3</sup><br/>
86
+ <em>Conference on AI Research, 2025</em><br/>
87
+ <a href="https://examplewebsite.com" title="Website" target="_blank" rel="nofollow" style="text-decoration: none;">🌎Website</a> |
88
+ <a href="https://examplecode.com" title="Dataset" target="_blank" rel="nofollow" style="text-decoration: none;">🧑‍💻Code</a> |
89
+ <a href="https://arxiv.org/examplepaper" title="aXiv" target="_blank" rel="nofollow" style="text-decoration: none;">📄arXiv</a> |
90
+ <a href="https://exampleleaderboard.com" title="Leaderboard" target="_blank" rel="nofollow" style="text-decoration: none;">🏆 Leaderboard (Coming Soon)</a><br>
91
  </p>
92
 
93
+ <img src="assets/img/logo.png" alt="Logo" width="400" height="auto" style="display:block; margin:auto;" />
 
 
94
  <p align='center' style="text-align:center;font-size:1.25em;">
95
+ <a align='center' style="text-decoration: none; color: gray">
96
+ Dataset is part of the <a href="">T* project</a></p>
97
+
98
+ <p align=center>
99
+ NOTE: Does Manling need Stanford Affiliation? <br>
100
+ NOTE: Fill in website url etc
 
 
 
 
 
 
 
 
 
 
101
  </p>
102
 
 
103
 
 
104
 
105
+ ## News
106
+
107
+ - **1/1/2025: Thrilled to announce T\* and LV-Haystack!**
108
+
109
+
110
+ ## Dataset Sample
111
+
112
+ ```python
113
+ {
114
+ 'vclip_id': '6338b73e-393f-4d37-b278-68703b45908c',
115
+ 'question_id': 10,
116
+ 'question': 'What nail did I pull out?',
117
+ 'answer': 'E',
118
+ 'frame_indexes': [5036, 5232],
119
+ 'choices': {
120
+ 'A': 'The nail from the front wheel fender',
121
+ 'B': 'The nail from the motorcycle battery compartment',
122
+ 'C': 'The nail from the left side of the motorcycle seat',
123
+ 'D': 'The nail from the rearview mirror mount',
124
+ 'E': 'The nail on the right side of the motorcycle exhaust pipe'
125
+ },
126
+ 'video_metadata': {
127
+ 'CLIP-reference-interval': [180.0, 240.0], # Time interval of the video clip
128
+ 'frame_count': 14155, # Total number of frames in the video
129
+ 'frame_rate': 30.0, # Frame rate of the video
130
+ 'duration': 471.8333435058594, # Duration of the video in seconds
131
+ 'resolution': '454x256', # Original resolution of the video
132
+ 'frame_dimensions': None, # Frame dimensions (if available)
133
+ 'codec': 'N/A', # Codec used for the video (not available here)
134
+ 'bitrate': 0, # Bitrate of the video
135
+ 'frame_dimensions_resized': [340, 256], # Resized frame dimensions
136
+ 'resolution_resized': '340x256', # Resized resolution
137
+ 'video_id': 'b6ae365a-dd70-42c4-90d6-e0351778d991' # Unique video identifier
138
+ }
139
+ }
140
+ ```
141
+
142
+ ## Usage
143
+
144
+ ```python
145
+ from datasets import load_dataset
146
+ dataset = load_dataset("LVHaystack/LongVideoHaystack")
147
+ print(dataset)
148
+ ```
149
+ ```bash
150
+ >>> DatasetDict({
151
+ train: Dataset({
152
+ features: ['vclip_id', 'question_id', 'question', 'answer', 'frame_indexes', 'choices', 'video_metadata'],
153
+ num_rows: 11218
154
+ })
155
+ test: Dataset({
156
+ features: ['vclip_id', 'question_id', 'question', 'answer', 'frame_indexes', 'choices', 'video_metadata'],
157
+ num_rows: 3874
158
+ })
159
+ })
160
+ ```
161
+
162
+
163
+
164
 
165
  ## Abstract
166
 
 
170
 
171
  <img src="[[STATISTICS_IMAGE_LINK]]" alt="image description" width="850" height="200">
172
 
 
 
 
 
 
 
 
 
 
 
 
 
173
 
 
174
 
175
  ## Dataset Organization
176
 
assets/img/logo.png ADDED