We generated Throwing Action dataset including following six categories for generic anomaly detection model.
- Bicycle Safe
- Bicycle Dangerous
- Car Safe
- Car Dangerous
- Pedestrian Safe
- Pedestrian Dangerous
Overall, the generated dataset consists of 130 normal videos without throwing anomalies and 271 anomalous videos divided over all six anomaly classes. All videos have a resolution of 320x240 pixels which is compatible with well-known UCF-Crime dataset. One of the challenges in our dataset is that throwing anomaly videos can contain multiple throwing actions up to a maximum of ten.
At this moment, complete dataset is not available yet because of the data privacy process. Please contact us to be able to reach the dataset.
