---
license: cc0-1.0
task_categories:
- video-classification
language:
- en
pretty_name: ucf_crime
configs:
- config_name: anomaly
  data_files:
    - split: train
      path: "data/anomaly-train-*.tar" # 1610 samples (1081 + 24 + 120 + 118 + 125 + 142)
    - split: test
      path: "data/anomaly-test-*.tar" # 290 samples (119 + 4 + 48 + 50 + 43 + 26)
- config_name: event1
  data_files:
    - split: train
      path:
        - "data/*-event-train-*.tar" # common 28 samples (anomaly-train:anomaly-test=24:4)
        - "data/*-event-test-2-*.tar"
        - "data/*-event-test-3-*.tar"
        - "data/*-event-test-4-*.tar"
    - split: test
      path: "data/*-event-test-1-*.tar" # 168 samples (anomaly-train:anomaly-test=120:48)
- config_name: event2
  data_files:
    - split: train
      path:
        - "data/*-event-train-*.tar"
        - "data/*-event-test-1-*.tar"
        - "data/*-event-test-3-*.tar"
        - "data/*-event-test-4-*.tar"
    - split: test
      path: "data/*-event-test-2-*.tar" # 168 samples (anomaly-train:anomaly-test=118:50)
- config_name: event3
  data_files:
    - split: train
      path:
        - "data/*-event-train-*.tar"
        - "data/*-event-test-1-*.tar"
        - "data/*-event-test-2-*.tar"
        - "data/*-event-test-4-*.tar"
    - split: test
      path: "data/*-event-test-3-*.tar" # 168 samples (anomaly-train:anomaly-test=125:43)
- config_name: event4
  data_files:
    - split: train
      path:
        - "data/*-event-train-*.tar"
        - "data/*-event-test-1-*.tar"
        - "data/*-event-test-2-*.tar"
        - "data/*-event-test-3-*.tar"
    - split: test
      path: "data/*-event-test-4-*.tar" # 168 samples (anomaly-train:anomaly-test=142:26)
size_categories:
- 10M<n<100M
---


# UCF-Crime Dataset (WebDataset Format)

이 데이터셋은 [UCF-Crime](https://www.crcv.ucf.edu/projects/real-world/) 데이터셋을 Hugging Face `datasets` 라이브러리의 스트리밍 기능을 위해 WebDataset(TAR Sharding) 형식으로 변환한 버전입니다.
100GB 이상의 대규모 비디오 데이터를 효율적으로 로드할 수 있습니다.

## Dataset Description
- **Homepage:** [CRCV UCF-Crime](https://www.crcv.ucf.edu/projects/real-world/)
- **Paper:** [Real-world Anomaly Detection in Surveillance Videos](https://arxiv.org/abs/1801.04264)


```python
from datasets import load_dataset

# Anomaly Detection 설정 로드
dataset = load_dataset("your_id/ucf_crime", "anomaly_detection", split="train", streaming=True)

for sample in dataset:
    video_path = sample["mp4"]
    label = sample["json"]["event"]
    print(f"Video loaded: {video_path}, Label: {label}")
    break
```

```
@InProceedings{Sultani_2018_CVPR,
    author={Sultani, Waqas and Chen, Chen and Shah, Mubarak},
    title={Real-World Anomaly Detection in Surveillance Videos},
    booktitle={The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    month={June},
    year={2018},
}
```