CosmicAstro/XD-Violence-X3D-Features
--- task_categories: - video-classification - anomaly-detection tags: - xd-violence - x3d - video-anomaly-detection - spatio-temporal-features - computer-vision - surveillance - violence-detection pretty_name: XD-Violence X3D Features --- # XD-Violence X3D Features ## Dataset Description This repository contains pre-extracted **X3D spatio-temporal features** from the XD-Violence dataset. The features are intended for research in **video anomaly detection (VAD)**, **weakly… See the full description on the dataset page: https://huggingface.co/datasets/CosmicAstro/XD-Violence-X3D-Features.
036
---
task_categories:
- video-classification
- anomaly-detection
tags:
- xd-violence
- x3d
- video-anomaly-detection
- spatio-temporal-features
- computer-vision
- surveillance
- violence-detection
pretty_name: XD-Violence X3D Features
---
# XD-Violence X3D Features
## Dataset Description
This repository contains pre-extracted **X3D spatio-temporal features**
from the XD-Violence dataset.
The features are intended for research in **video anomaly detection
(VAD)**, **weakly supervised video anomaly detection (WSVAD)**, and
violence/anomaly recognition from surveillance and real-world videos.
This repository contains extracted features rather than the original
raw videos.
## Source Dataset
The features were extracted from videos belonging to the
**XD-Violence** dataset, a large-scale audio-visual dataset designed
for violence detection and video anomaly detection research.
## Feature Extraction
The videos were processed using an **X3D** spatio-temporal
convolutional network to obtain compact video representations.
The resulting features can be used as input to downstream temporal
models and anomaly detection architectures.
## Repository Contents
Current repository statistics:
- **Dataset:** XD-Violence
- **Feature extractor:** X3D
- **Feature files:** 4,750
- **Approximate size:** 5.5 GB
## Intended Use
These features can be used for research involving:
- Video anomaly detection
- Weakly supervised video anomaly detection
- Violence detection
- Spatio-temporal representation learning
- Surveillance video understanding
- Temporal anomaly detection
- Reproducible VAD experiments
## Data Processing
The original video sequences were processed using X3D to generate
spatio-temporal feature representations.
The extracted representations are organized for use by downstream
video anomaly detection and temporal modeling pipelines.
## Important Note
This repository contains **pre-extracted X3D features and not the
original XD-Violence raw videos**.
Users should obtain the original XD-Violence dataset from its
appropriate source and comply with its terms of use.
This repository does not claim ownership of the original XD-Violence
videos.
## Citation
If you use these features in academic research, please cite the
original XD-Violence dataset and the X3D work used to generate the
features.
## License
The licensing and usage conditions of the original XD-Violence dataset
apply to the underlying source data.
Please consult the original dataset terms before using or
redistributing the data.