Hugging Face's logo

Datasets:
constanceCM
/
OpenTAD_Save

Dataset card Files
xet
Community
OpenTAD_Save / OpenTAD /tools /prepare_data /fineaction /README.md
constanceCM's picture
constanceCM
upload
8aa674c
preview code
|
raw
history blame contribute delete
2.14 kB

Data Preparation for FineAction-1.3

Download Annotations

You can download the annotation by

bash download_annotation.sh

The annotation should be in data/fineaction/annotations/.

Download Pre-extracted Features

Please put the downloaded feature under the path: data/fineaction/features/.

We provide the following pre-extracted features:

Feature Url Backbone Feature Extraction Setting
videomae_h InternVideo Repo VideoMAE-H-K700-16x4x1 snippet_stride=16, clip_length=16, frame_interval=1
videomaev2_g VideoMAEv2 Repo VideoMAEv2-g-K710-16x4x1 snippet_stride=16, clip_length=16, frame_interval=1
internvideo2_1b Official Repo InternVideo2-6B snippet_stride=4, clip_length=16, frame_interval=1

Download Raw Videos

Please put the downloaded video under the path: data/fineaction/raw_data/.

You can download the raw video from official website.

Citation 

@article{liu2022fineaction,
  title={Fineaction: A fine-grained video dataset for temporal action localization},
  author={Liu, Yi and Wang, Limin and Wang, Yali and Ma, Xiao and Qiao, Yu},
  journal={IEEE Transactions on Image Processing},
  year={2022},
  publisher={IEEE}
}