| --- |
| license: cc-by-4.0 |
| task_categories: |
| - video-classification |
| tags: |
| - action-recognition |
| - temporal-segmentation |
| - video |
| - computer-vision |
| size_categories: |
| - 1K<n<10K |
| --- |
| |
| # The Breakfast Actions Dataset |
|
|
| **Official dataset release by Serre Lab, Brown University.** |
|
|
| A comprehensive dataset of breakfast preparation activities performed by 52 different individuals in 18 different kitchens (~77 hours of video, >4 million frames). |
|
|
| ## Dataset Specifications |
|
|
| - **Participants**: 52 individuals |
| - **Locations**: 18 kitchens |
| - **Total recording**: ~77 hours of video (>4 million frames) |
| - **Resolution**: 320x240 pixels |
| - **Frame rate**: 15 fps |
| - **Activities**: 10 breakfast preparation activities, 48 fine-grained action units |
|
|
| ## Files |
|
|
| - **demo_bundle.rar** — MATLAB demo for action recognition |
| - **BreakfastII_15fps_qvga_sync.tar.gz** — Video data |
| - **breakfast_data.tar.gz** — Dataset annotations |
| - **segmentation_coarse.tar.gz** — Coarse temporal segmentation |
| - **segmentation_fine.tar.gz** — Fine temporal segmentation |
| |
| ## Citation |
| |
| ```bibtex |
| @inproceedings{kuehne2014language, |
| title={The Language of Actions: Recovering the Syntax and Semantics of Goal-Directed Human Activities}, |
| author={Kuehne, Hilde and Arslan, Ali and Serre, Thomas}, |
| booktitle={CVPR}, |
| year={2014} |
| } |
| ``` |
| |
| ## Links |
| |
| - [Official website](https://serre.lab.brown.edu/breakfast-actions-dataset.html) |
| - [GitHub: HTK Action Recognition](https://github.com/hildekuehne/HTK_actionRecognition) |
| |
