--- dataset_info: features: - name: vclip_id dtype: string - name: question_id dtype: int64 - name: question dtype: string - name: answer dtype: string - name: frame_indexes sequence: int64 - name: choices struct: - name: A dtype: string - name: B dtype: string - name: C dtype: string - name: D dtype: string - name: E dtype: string - name: video_metadata struct: - name: CLIP-reference-interval-clip sequence: float64 - name: CLIP-reference-interval-video sequence: float64 - name: bitrate dtype: int64 - name: codec dtype: string - name: frame_dimensions sequence: int64 - name: frame_dimensions_resized sequence: int64 - name: frame_rate dtype: float64 - name: resolution dtype: string - name: resolution_resized dtype: string - name: vclip_duration dtype: float64 - name: vclip_frame_count dtype: int64 - name: vclip_interval_in_video sequence: float64 - name: video_duration dtype: float64 - name: video_frame_count dtype: int64 - name: video_id dtype: string splits: - name: train num_bytes: 5358616 num_examples: 11218 - name: test num_bytes: 1977870 num_examples: 3874 download_size: 2168577 dataset_size: 7336486 configs: - config_name: default data_files: - split: train path: data/train-* - split: test path: data/test-* ---

LV-Haystack: Temporal Search for Long-Form Video Understanding

Jinhui Ye¹, Zihan Wang², Haosen Sun², Keshigeyan Chandrasegaran¹,
Zane Durante¹, Cristobal Eyzaguirre¹, Yonatan Bisk³, Juan Carlos Niebles¹, Ehsan Adeli¹,
Li Fei-Fei¹, Jiajun Wu¹, Manling Li²
Stanford University¹, Northwestern University², Carnegie Mellon University³
Dataset is part of the T* project
🌎Website | 🧑‍💻Code | 📄arXiv | 🏆 Leaderboard (Coming Soon)

#### Dataset Sample ```python { 'vclip_id': '6338b73e-393f-4d37-b278-68703b45908c', 'question_id': 10, 'question': 'What nail did I pull out?', 'answer': 'E', 'frame_indexes': [5036, 5232], # the keyframe indexes 'choices': { 'A': 'The nail from the front wheel fender', 'B': 'The nail from the motorcycle battery compartment', 'C': 'The nail from the left side of the motorcycle seat', 'D': 'The nail from the rearview mirror mount', 'E': 'The nail on the right side of the motorcycle exhaust pipe' }, 'video_metadata': { 'CLIP-reference-interval-vclip': [180.0, 240.0], # Time interval of the "vclip" that is considered to be important by CLIP. this is calculated by (CLIP-reference-interval-video - vclip-interval-in-video[0]) 'CLIP-reference-interval-video': [180.0, 240.0], # Time interval of the "video" that is considered to be important by CLIP. This is originally from the **Ego4D dataset**, used in our work for annotators to quickly locate in the video. 'vclip_interval_in_video': [0.0, 480.06667277018227], # the vclip start and end second, i.e., for [a, b], the vclip starts at the a second of the video, ends at the b second of the video 'frame_count': 14155, # Total number of frames in the video 'frame_rate': 30.0, # Frame rate of the video 'duration': 471.8333435058594, # Duration of the video that are valid and unbroken, in seconds 'resolution': '454x256', # Original resolution of the video 'frame_dimensions': None, # Frame dimensions (if available) 'codec': 'N/A', # Codec used for the video (if available) 'bitrate': 0, # Bitrate of the video (if available) 'frame_dimensions_resized': [340, 256], # Resized frame dimensions 'resolution_resized': '340x256', # Resized resolution 'video_id': 'b6ae365a-dd70-42c4-90d6-e0351778d991' # Unique video identifier } } ``` #### Dataset exploration add hyperlink to demo #### Dataset Usage ```python from datasets import load_dataset dataset = load_dataset("LVHaystack/LongVideoHaystack") print(dataset) ``` ```bash >>> DatasetDict({ train: Dataset({ features: ['vclip_id', 'question_id', 'question', 'answer', 'frame_indexes', 'choices', 'video_metadata'], num_rows: 11218 }) test: Dataset({ features: ['vclip_id', 'question_id', 'question', 'answer', 'frame_indexes', 'choices', 'video_metadata'], num_rows: 3874 }) }) ``` #### Video Source Download TODO: We plan to provide a script of how to download a subset from [Ego4d](https://ego4d-data.org/). For now, you can refer to their official guide [here](https://github.com/facebookresearch/Ego4d/tree/main/ego4d/cli). Your code would be look like the follows: ```bash pip install ego4d ego4d --output_directory=your_path/videos/ \ --datasets full_scale annotations \ --metadata \ --video_uid_file video_uids.txt python process_videos_to_clips.py # TODO ``` Please find [video_uid.txt](https://huggingface.co/datasets/LVHaystack/LongVideoHaystack/blob/main/video_uid.txt) in our repo, or you can generate it by: ```python import datasets metadata = datasets.load_dataset("LVHaystack/LongVideoHaystack-metadata")["metadata"] with open("video_uids.txt", "w") as file: for video_id in list(set(metadata['video_id'])): file.write(video_id + " ") ``` then, you need to transform them to video clips: ```python ``` #### Dataset Statistics Summary | **Metric** | **Total** | **Train** | **Test** | |-------------------------------|--------------|-------------|-------------| | **Video Statistics** | | | | | Total Videos | **988** | **744** | **244** | | Total Video Duration (hr) | 423.3 | 322.2 | 101.0 | | Avg. Video Duration (min) | 25.7 | 26.0 | 24.8 | | **Clip Statistics** | | | | | Total Video Clips | **1,324** | **996** | **328** | | Total Video Clip Duration (hr) | 180.4 | 135.3 | 45.0 | | Avg. Video Clip Duration (sec) | 8.2 | 8.2 | 8.2 | | **Frame Statistics** | | | | | Total Frames (k) | **45,700** | **34,800** | **10,900** | | Avg. Frames per Video (k) | 46.3 | 46.8 | 44.7 | | Ratio of Keyframe / Frame (‰) | 0.62 | 0.59 | 0.71 | | **QA Statistics** | | | | | Total QA Pairs | **15,092** | **11,218** | **3,874** | | Avg. QA Pair per Video | 15.3 | 15.1 | 15.9 | | Avg. QA Pair per Clip | 11.4 | 11.3 | 11.8 | | Avg. Keyframes per Question | 1.88 | 1.84 | 2.01 | #### Evaluation scripts Please refer to [./eval.py](https://huggingface.co/datasets/LVHaystack/LongVideoHaystack/blob/main/eval.py). #### Contact - Jinhui Ye: jinhuiyes@gmail.com - Zihan Wang: zihanw@u.northwestern.edu (datasets) - Haosen Sun: haosensun2026@u.northwestern.edu - Keshigeyan Chandrasegaran: keshik@stanford.edu - Manling Li: manling.li@northwestern.edu #### Citation ```bibtex @misc{tstar, title={Re-thinking Temporal Search for Long-Form Video Understanding}, author={Jinhui Ye and Zihan Wang and Haosen Sun and Keshigeyan Chandrasegaran and Zane Durante and Cristobal Eyzaguirre and Yonatan Bisk and Juan Carlos Niebles and Ehsan Adeli and Li Fei-Fei and Jiajun Wu and Manling Li}, year={2025}, eprint={2501.TODO}, archivePrefix={arXiv}, primaryClass={cs.LG} } ``` Website template borrowed from [HourVideo](https://huggingface.co/datasets/HourVideo/HourVideo).