imageomics/kabr-methodology

Dataset Card for kabr-tools Methodology Dataset

Dataset Details

A curated collection of CSV and XML files describing time-budget data, focal observations, scan samples, and object-detection annotations for African ungulates—including Grevy’s zebras, plains zebras, and giraffes—recorded both from the ground and from drones. This dataset complements the original KABR Mini-Scene Dataset, providing ground-based sampling to correspond with a subset of the published mini-scenes. Specifically designed to compare Kenyan animal behavior observation methods, this dataset supports:

• Object detection benchmarks (PASCAL-VOC XML).
• Behavior classification and time-budget analyses from per-individual CSV logs.
• Methodological comparisons between focal, scan, and drone-based sampling.

Figure 1. Time budget comparison between drone-based behavior classification (bottom) and manual field focal observations (top).

Dataset Structure

kabr-methodology
├── focalvsdrone/
│   ├── focal_drone_df_12_01_23_female_grevy.csv
│   ├── focal_drone_df_16_01_23_thick_neck_stripes.csv
│   ├── focal_drone_df_16_01_23_white_female.csv
│   ├── focal_drone_df_17_01_23_scar_cleaned.csv
│   └── readme.txt
└── scanvsfocal/
    ├── giraffe/
    │   ├── giraffe_focal1-female.csv
    │   ├── giraffe_focal1-male.csv
    │   ├── giraffe_focal2-white_neck_male.csv
    │   ├── Scan_giraffe_group_1.csv
    │   ├── Scan_giraffe_group_2.csv
    ├── grevys/
    │   ├── Focal_grevys_group_1_morning_01_11.csv
    │   ├── Focal_grevys_group_2_a_01_11.csv
    │   ├── Focal_grevys_group_2_b_01_11.csv
    │   ├── Focal_grevys_group_2_c_01_11.csv
    │   ├── Scan_grevys_group_1_morning_01_11.csv
    │   ├── Scan_grevys_group_2_a_01_11.csv
    │   ├── Scan_grevys_group_2_b_01_11.csv
    │   ├── Scan_grevys_group_2_c_01_11.csv
    └── plain/
        ├── plain_Focal-group_1.csv
        ├── plain_Focal-group_2.csv
        ├── plain_Scan-group_1.csv
        └── plain_Scan-group_2.csv

Data Instances and Files (by Folder/Task)

focalvsdrone/

Description:

Paired focal observation and synchronized drone log for four individual zebras. Each CSV contains time-aligned behavior states from ground focal sampling and drone footage, enabling direct comparison of behavioral data collected via the two methods. Note that the zebras in these files are given descriptive nicknames to distinguish them.

Files (CSVs):

• focal_drone_df_12_01_23_female_grevy.csv — Paired focal observation and synchronized drone log for a female Grevy's zebra on 12-01-2023.
  • 'Zebra A': This zebra is found in video DJI_0997, miniscenes 12, 20, 33, 39, 62, and DJI_0998, miniscene 1.
• focal_drone_df_16_01_23_thick_neck_stripes.csv — Paired focal vs drone for ID "thick_neck_stripes" on 16-01-2023.
  • 'Zebra B': This zebra is found in DJI_0001, mini-scene 47 and DJI_0002, mini-scene 5 and 9.
• focal_drone_df_16_01_23_white_female.csv — Paired focal vs drone for "white_female" on 16-01-2023.
  • 'Zebra C': This zebra is found in DJI_0001, miniscene 8,9,11,31,44.
• focal_drone_df_17_01_23_scar_cleaned.csv — Cleaned paired focal vs drone trace for "scar" on 17-01-2023.
  • 'Zebra D':This zebra is found in DJI_0008, mini-scene 40 and 41, and DJI_010 mini-scene 5.

CSV Contents:

Column	Description
frame	Sequential frame number in the processed dataset
behavior	Behavior label from drone video annotation in CVAT
video_frame	Original frame number from the source video file
time-date	Timestamp (YYYY-MM-DD HH:MM:SS in local time)
focal_behavior	Behavior recorded from focal animal observation

kabr-methodology/scanvsfocal/

Description:

Paired scan sample and focal observation logs for groups of giraffes, Grevy’s zebras, and plains zebras. Each CSV contains time-aligned behavior states from group-level scan sampling and individual focal sessions, enabling direct comparison of behavioral data collected via the two methods, exported from the AnimalBehaviourPro app.

Files (CSVs):

Giraffes:

• giraffe_focal1-female.csv — Giraffe focal log (behavior states over time).
• giraffe_focal1-male.csv — Giraffe focal log (behavior states over time).
• giraffe_focal2-white_neck_male.csv — Giraffe focal log (behavior states over time).
• Scan_giraffe_group_1.csv — Group-level scan sample for giraffes (group 1).
• Scan_giraffe_group_2.csv — Group-level scan sample for giraffes (group 2).

Grevy's Zebras:

• Focal_grevys_group 1 morning 01_11.csv — Group 1 focal session (AM), 01-11.
• Focal_grevys_group 2_a 01_11.csv — Group 2 focal session (individual a), 01-11.
• Focal_grevys_group 2_b 01_11.csv — Group 2 focal session (individual b), 01-11.
• Focal_grevys_group 2_c 01_11.csv — Group 2 focal session (individual c), 01-11.
• Scan_grevys_group 1 morning 01_11.csv — Group 1 scan sample (AM), 01-11.
• Scan_grevys_group 2_a 01_11.csv — Group 2 scan sample (individual a), 01-11.
• Scan_grevys_group 2_b 01_11.csv — Group 2 scan sample (individual b), 01-11.
• Scan_grevys_group 2_c 01_11.csv — Group 2 scan sample (individual c), 01-11.

Plains Zebras:

• plain_Focal-group_1.csv — Plains zebra focal log (group 1).
• plain_Focal-group_2.csv — Plains zebra focal log (group 2).
• plain_Scan-group_1.csv — Plains zebra group scan (group 1).
• plain_Scan-group_2.csv — Plains zebra group scan (group 2).

CSV Contents:

Each of the above listed CSVs has the columns indicated below, where each row indicates a new behavior observation for a single individual. The one exception is Scan_giraffe_group_2.csv, which resulted from testing a different observation type; differences with that CSV are described in Scan all behavior type, after the table.

Column	Description
Date_ymd	Date of observation (YYYY-MM-DD format)
Time_Absolute_hms	Absolute time of day (local) when behavior was recorded (HH:MM:SS)
Time_Relative_hms	Time relative to start of observation session (HH:MM:SS)
Time_Relative_s	Time relative to start of observation session in seconds
Time_Lag_s	Time lag between current and previous event in seconds (i.e., time to log behavior in app)
Duration_s	Duration of the behavioral state in seconds
Observation_Name	Type of observation method ("AdLib" for ad libitum sampling, most observations)
Log_File_Name	Source log file name -- original name of this file
Actor	Individual animal being observed (e.g., "Male w/ scar on neck")
Behavior	Observed behavior (e.g., "Stand", "Walk")
Receiver	Target of social behaviors (empty for non-social behaviors)
Modifier_1 through Modifier_6	Additional behavioral modifiers (mostly empty)
Other_Modifiers	Additional modifiers not captured in numbered fields (mostly empty)
Event_Type	Whether this marks the start or stop of a behavioral state ("State start" or "State stop")
Observer_ID	Unique identifier for the observer (e.g., "1234")
Notes	Additional notes about the observation (empty in this dataset)
Coding_Scheme	Behavioral coding scheme used ("Zebra")

Scan all behavior type

Scan_giraffe_group_2.csv is a product of the "Scan all behavior" observation method. Instead of producing one line per behavior per subject, this observation method produced one line per observation for all individuals and their behaviors (at a particular time). As a result, the Actor, Behavior, Receiver, Modifier_1, through Modifier_6, Other_Modifiers, and Event_Type columns are replaced by Group_Size, Activity, Subjects, Brown male_Behaviour, Brown male_Receiver, Brown male_Modifiers, White male_Behaviour, White male_Receiver, and White male_Modifiers, where "White" and "Brown" males are the two giraffes being observed. The Activity column is filled in when both are performing the same behavior (e.g., "Walking"), in which case their individual behaviors are also both filled in. Note that in this CSV behavior is indicated "Behaviour", but the others have the column "Behavior".

Data Splits

The dataset ships as a single corpus. Create custom splits by video ID, individual ID, or survey date to control cross-view generalization (e.g., ground vs drone).

Dataset Creation

Curation Rationale

This dataset was curated for the purposes of

(i) quantifying the strengths and weaknesses of ground- vs drone-based behavioral sampling;

(ii) providing detection and coarse behavior benchmarks in the wild.

Collection & Processing

Ground focal/scan samples recorded by trained observers; DJI drone footage (~30 m AGL); CVAT for bounding boxes; custom scripts for ethogram coding; standardized CSV/XML exports. Field scan and focal sampling collected with AnimalBehaviourPro App.

Source Data Producers

Field teams at Mpala Research Centre, Laikipia (Kenya), worked with this group to collect the animal behavior observations.

Considerations for Using the Data

Bias • Risks • Limitations

• Class imbalance (Grevy’s zebras dominate) may bias detectors/behavior models.
• Daytime-only footage and observations limit potential for generalization to low-light/night conditions.
• Coarse behavior labels omit fine motor actions.
• Recommendations: Use stratified sampling or class weighting; evaluate with cross-view splits (ground vs drone).

Licensing

This compilation is dedicated to the public domain (released under CC0 1.0) for the benefit of scientific pursuits.

Citation

If you make use of this dataset for your research, please cite both it and the original KABR mini-scene as below:

Datasets:

@misc{kline2025kabr-tools-methodology,
  author    = {Jenna Kline and Maksim Kholiavchenko and Michelle Ramirez and Samuel Stevens and
               Reshma Ramesh Babu and Namrata Banerji and Elizabeth Campolongo and Matthew Thompson and
               Nina Van Tiel and Jackson Miliko and Isla Duporge and Neil Rosser and Chuck Stewart and
               Tanya Berger-Wolf and Daniel Rubenstein},
  title     = {kabr-tools Methodology Dataset (Revision 67de7c6)},
  year      = {2025},
  url       = {https://huggingface.co/datasets/imageomics/kabr-methodology},
  publisher = {Hugging Face},
  doi       = {10.57967/hf/6791}
}

@misc{KABR_Data,
  author = {Kholiavchenko, Maksim and Kline, Jenna and Ramirez, Michelle and Stevens, Sam and Sheets, Alec and Babu, Reshma and Banerji, Namrata and Campolongo, Elizabeth and Thompson, Matthew and Van Tiel, Nina and Miliko, Jackson and Bessa, Eduardo and Duporge, Isla and Berger-Wolf, Tanya and Rubenstein, Daniel and Stewart, Charles},
  title = {KABR: In-Situ Dataset for Kenyan Animal Behavior Recognition from Drone Videos},
  year = {2023},
  url = {https://huggingface.co/datasets/imageomics/KABR},
  doi = {10.57967/hf/1010},
  publisher = {Hugging Face}
}

kabr-tools manuscript (pre-print)

@misc{kline2025kabrtoolsautomatedframeworkmultispecies,
      title={kabr-tools: Automated Framework for Multi-Species Behavioral Monitoring}, 
      author={Jenna Kline and Maksim Kholiavchenko and Samuel Stevens and Nina van Tiel and Alison Zhong and Namrata Banerji and Alec Sheets and Sowbaranika Balasubramaniam and Isla Duporge and Matthew Thompson and Elizabeth Campolongo and Jackson Miliko and Neil Rosser and Tanya Berger-Wolf and Charles V. Stewart and Daniel I. Rubenstein},
      year={2025},
      eprint={2510.02030},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2510.02030}, 
}

KABR mini-scene Paper:

@inproceedings{kholiavchenko2024kabr,
  title={KABR: In-Situ Dataset for Kenyan Animal Behavior Recognition from Drone Videos},
  author={Kholiavchenko, Maksim and Kline, Jenna and Ramirez, Michelle and Stevens, Sam and Sheets, Alec and Babu, Reshma and Banerji, Namrata and Campolongo, Elizabeth and Thompson, Matthew and Van Tiel, Nina and Miliko, Jackson and Bessa, Eduardo and Duporge, Isla and Berger-Wolf, Tanya and Rubenstein, Daniel and Stewart, Charles},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={31-40},
  year={2024}
}

Acknowledgements

This work was supported by the Imageomics Institute, which is funded by the US National Science Foundation's Harnessing the Data Revolution (HDR) program under Award #2118240 (Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning). Additional support was also provided by the AI Institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE), which is funded by the US National Science Foundation under Award #2112606. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

We thank the field experts at Mpala Research Centre for their support with data collection and logistics.

The data was gathered at the Mpala Research Centre in Kenya, in accordance with Research License No. NACOSTI/P/22/18214. The data collection protocol adhered strictly to the guidelines set forth by the Institutional Animal Care and Use Committee under permission No. IACUC 1835F.

Dataset Card Authors

Jenna Kline and Elizabeth Campolongo

Contact

Questions or benchmark submissions? Open an issue on the GitHub repo: https://github.com/Imageomics/kabr-tools

Dataset Card Contact

kline dot 377 at osu dot edu