|
|
--- |
|
|
annotations_creators: |
|
|
- expert-generated |
|
|
language: |
|
|
- und |
|
|
multilingual: false |
|
|
pretty_name: HMDB51 |
|
|
task_categories: |
|
|
- video-classification |
|
|
license: other |
|
|
dataset_creators: |
|
|
- Serre Lab at Brown University |
|
|
paperswithcode_id: hmdb51 |
|
|
tags: |
|
|
- human-action |
|
|
- video |
|
|
- benchmark |
|
|
--- |
|
|
|
|
|
# HMDB51 Dataset |
|
|
|
|
|
**HMDB51** is a large human action recognition dataset consisting of 6,766 video clips (.avi) distributed across 51 action categories (e.g., jumping, drinking, kicking, etc). It is commonly used as a benchmark for video classification and action recognition tasks. |
|
|
|
|
|
## 📦 Dataset Source |
|
|
|
|
|
This dataset was originally collected and published by the **Serre Lab at Brown University**. |
|
|
|
|
|
🔗 [Serre Lab – Brown University](https://serre-lab.clps.brown.edu/) |
|
|
|
|
|
## 📁 Contents |
|
|
|
|
|
- 6,766 video clips |
|
|
- 51 action categories |
|
|
- Includes training/testing splits for benchmarking |
|
|
|
|
|
## 📚 Citation |
|
|
|
|
|
If you use this dataset, please cite: |
|
|
|
|
|
> Kuehne, H., Jhuang, H., Garrote, E., Poggio, T., & Serre, T. (2011). HMDB: A large video database for human motion recognition. In Proceedings of the IEEE International Conference on Computer Vision (ICCV). |
|
|
|
|
|
|
