Skip to content
aistaircPublic

About

Data and code for the paper "FineBio: A Fine-Grained Video Dataset of Biological Experiments with Hierarchical Annotation"

Resources

Stars

28 stars

Watchers

3 watching

Forks

Repository files navigation

FineBio: A Fine-Grained Video Dataset of Biological Experiments with Hierarchical Annotation

This repository includes the download instuction and the code for the paper FineBio: A Fine-Grained Video Dataset of Biological Experiments with Hierarchical Annotation (IJCV, 2025).

teaser_v2

Download

Usage is limited to non-commercial research/development use. Users need to sign the license form to access the videos, metadata and annotation. Please sign to the license agreement and submit via this form.

Link to the dataset and required credentials will be sent by e-mail after approval.

26/09/10 Update: We have updated the license agreement. Please use the latest version for application.
25/12/15 Update: We have updated the download links to the pre-trained models.
24/09/12 Update: The object detection annotation (finebio_coco_annotations.zip) has been updated because there were missing annotation. Please re-download it if you are using the older version.

Dataset

Name convention

Video and annotation files are named by the following rules:

# First-person view
P<participant_id (1-32)>_<protocol_id (1-7)>_<take_id>.mp4
# Third-person view
P<participant_id (1-32)>_<protocol_id (1-7)>_<take_id>_T<camera_id (1-5)>.mp4
# Atomic operation annotation
P<participant_id (1-32)>_<protocol_id (1-7)>_<take_id>.txt
# Object detection images
P<participant_id (1-32)>_<protocol_id (1-7)>_<take_id>_<frame_num>.jpg
P<participant_id (1-32)>_<protocol_id (1-7)>_<take_id>__T<camera_id (1-5)>_frame_num>.jpg

Pre-trained Models

Please refer to the README for each benchmark for details.

Step segmentation

Backbone Model Acc Edit F1@10 F1@25 F1@50 F1@75 Weights
I3D MS-TCN++ 90.2 96.7 97.4 96.7 93.5 73.4 Link
I3D ASFormer 87.2 94.8 94.2 92.7 86.5 67.0 Link

Atomic Operation Detection

Backbone Model mAP@0.3 0.4 0.5 0.6 0.7 Avg. Weights
I3D ActionFormer (multi-head + set classification) 45.2 41.7 36.5 28.4 18.7 34.1 Link

Object Detection

Model AP AP50 AP_manipulated AP_affected Weights
DINO 53.3 77.4 55.9 51.6 Link
Deformable DETR 56.1 78.5 64.0 58.8 Link

Manipulated/Affected Object Detection

Hand H + Manipulated H + M + Affected Weights
Left Hand 96.8 6.5 5.9 Link
Right Hand 94.5 22.2 10.7

Previous methods used in the implementation

Please cite the necessary methods if you used our baseline models.

Support

If you find any problem, please report to Takuma Yagi (takuma.yagi[at]aist.go.jp) or by this repository's issue.

Citation

Please cite our work if you have used our data or code:

@article{yagi2025finebio,
  title={Finebio: A fine-grained video dataset of biological experiments with hierarchical annotation},
  author={Yagi, Takuma and Ohashi, Misaki and Huang, Yifei and Furuta, Ryosuke and Adachi, Shungo and Mitsuyama, Toutai and Sato, Yoichi},
  journal={International Journal of Computer Vision},
  volume={133},
  pages={7352--7367},
  year={2025},
  publisher={Springer}
}

License

The code is licensed under the MIT License.

Frequently Asked Questions (FAQs)

  • Q: How long does it take for the application to be accepted?
  • A: We need to check the application manually. It will be typically accepted within two or three business days, but it may cause some delays around holidays.
  • Q: My request has been declined. What is the reason?
  • A: Please provide correct affiliation, title, name, and your purpose of use. Please provide your Signature in the PDF agreement.
  • Q: The download link has been expired.
  • A: Please re-submit the form so that we can publish new download link. Note that the download link will expire in two weeks.
  • Q: I'd like to use the dataset for training/evaluating commercial models. Do you offer commercial licensing?
  • A: We are currently preparing to offer paid license agreements. Please e-mail Takuma Yagi (takuma.yagi[at]aist.go.jp) for details.

About

Data and code for the paper "FineBio: A Fine-Grained Video Dataset of Biological Experiments with Hierarchical Annotation"

Resources

Stars

28 stars

Watchers

3 watching

Forks

Releases

Packages

Used by

Contributors

Languages