imageomics/kabr-datapalooza-2023-subset

Dataset Card for KABR Drone Wildlife Monitoring Dataset

This dataset consists of synchronized drone telemetry, camera metadata, behavior annotations, and drone status data collected during wildlife monitoring operations. The data was captured using drones equipped with cameras to observe animal behavior in natural habitats.

Dataset Details

This dataset integrates multiple streams of information collected during drone monitoring of wildlife in Kenya. It combines precise drone telemetry data (position, orientation, altitude), camera settings (ISO, shutter speed, focal length), wildlife annotations (bounding boxes, behavior classification), and drone system status information. The dataset was developed as part of research on integrating biological data into autonomous remote sensing systems for in situ imageomics, specifically focused on Kenyan animal behavior sensing with Unmanned Aerial Vehicles (UAVs). The dataset provides a comprehensive framework for analyzing animal behavior in correlation with drone positioning and camera settings, enabling research in fields such as wildlife monitoring, animal behavior analysis, and drone-based observation methodologies.

Supported Tasks and Leaderboards

This dataset supports several computer vision and behavioral analysis tasks:

1. Object detection and tracking of animals in drone footage
2. Behavior classification and analysis
3. Correlating animal behavior with drone positioning and movement
4. Optimizing drone flight patterns for wildlife observation
5. Camera parameter optimization for wildlife monitoring

Dataset Structure

The dataset is organized with the following structure:

/dataset/
    metadata.csv       # Main metadata file with all information

Data Instances

Each row in the metadata.csv file represents a single frame from a drone video with associated telemetry, annotations, and status information. The dataset contains [number] frames from [number] videos, collected between [dates].

Data Fields

The metadata file contains 87 columns organized into four main categories:

1. Camera Settings and Frame Information

• frame: Frame number in the video sequence
• id: Unique identifier for the frame
• iso: ISO setting of the camera
• shutter: Shutter speed value
• fnum: Aperture f-number
• ev: Exposure value
• ct: Color temperature
• color_md: Color mode
• focal_len: Focal length of the camera lens in mm
• dzoom_ratio: Digital zoom ratio
• isPhoto: Binary flag indicating if the frame is a photo
• isVideo: Binary flag indicating if the frame is from video
• video: Source video file name

2. Geo-location and Timing Data

• latitude_x, latitude_y: GPS latitude coordinates
• longitude_x, longitude_y: GPS longitude coordinates
• altitude: Altitude of the drone
• date_time_x, date_time_y, date_time: Timestamps
• ms, new_ms, time(millisecond): Millisecond timing information

3. Wildlife Annotation Data

• xtl, ytl, xbr, ybr: Bounding box coordinates (top-left x,y and bottom-right x,y)
• points: Polygon or keypoint annotations
• label: Class label of the annotated object
• source: Source of the annotation
• behaviour: Annotated behavior classification
• keyframe_x, keyframe_y: Keyframe indicators for tracking
• outside_x, outside_y: Flags indicating if the subject is outside the frame
• occluded_x, occluded_y: Flags indicating if the subject is occluded
• z_order_x, z_order_y: Z-order for overlapping annotations

4. Drone Status and Telemetry

• height_above_takeoff(feet): Drone height relative to takeoff point
• height_above_ground_at_drone_location(feet): Drone height above ground
• ground_elevation_at_drone_location(feet): Ground elevation at drone location
• altitude_above_seaLevel(feet): Altitude above sea level
• height_sonar(feet): Height measured by sonar
• speed(mph): Drone speed
• distance(feet): Distance from takeoff point
• mileage(feet): Total distance traveled
• satellites: Number of GPS satellites being used
• gpslevel: GPS signal strength
• voltage(v): Battery voltage
• max_altitude(feet), max_ascent(feet), max_speed(mph), max_distance(feet): Maximum values recorded
• xSpeed(mph), ySpeed(mph), zSpeed(mph): Speed components in x, y, z directions
• compass_heading(degrees): Compass heading
• pitch(degrees), roll(degrees): Drone orientation angles
• rc_elevator, rc_aileron, rc_throttle, rc_rudder: Remote control inputs
• rc_elevator(percent), rc_aileron(percent), rc_throttle(percent), rc_rudder(percent): Remote control inputs as percentages
• gimbal_heading(degrees), gimbal_pitch(degrees), gimbal_roll(degrees): Gimbal orientation
• battery_percent: Battery percentage remaining
• voltageCell1 through voltageCell6: Individual battery cell voltages
• current(A): Current draw in Amperes
• battery_temperature(f): Battery temperature in Fahrenheit
• flycStateRaw: Raw flight controller state
• flycState: Human-readable flight controller state
• message: Status or event message

Dataset Creation

Curation Rationale

This dataset was specifically curated to combine behavior annotations with corresponding drone telemetry to analyze which flight features produced useful data for behavior analysis. By integrating these different data streams, researchers can identify optimal drone flight patterns, heights, speeds, and camera configurations that maximize the quality of behavioral data while minimizing disturbance to wildlife. This integration supports research in wildlife monitoring and behavior analysis using drone technology, with the goal of developing improved protocols for wildlife observation that balance data quality with animal welfare considerations.

Source Data

Data Collection and Processing

Data was collected from 6 January 2023 through 21 January 2023 at the Mpala Research Centre in Kenya under a Nacosti research license. The team used DJI Mavic 2S drones equipped with cameras to record 5.4K resolution videos (5472 x 3078 pixels) from varying altitudes and distances of 10 to 50 meters from the animals. The distance was determined by circumstances and safety regulations to ensure both quality data collection and minimal wildlife disturbance.

Frame extraction was performed using CVAT, and behavior annotations were added using [annotation tool/software] by [annotators]. Telemetry data was synchronized with video frames using [method/software].

Who are the source data producers?

The dataset was collected by the Imageomics team as part of the Kenyan Animal Behavior Remote sensing (KABR) project. The drone operations were conducted by licensed drone operators and researchers with appropriate permits for wildlife observation in Kenya.

Annotations

Annotation process

Please refer to the KABR dataset and associated paper for details on the annotation process.

Personal and Sensitive Information

This dataset does not contain personal information. Location data has been [method for handling sensitive wildlife location data, if applicable] to protect vulnerable or endangered species.

Considerations for Using the Data

Bias, Risks, and Limitations

• Sampling bias: Data collection was limited to [specific conditions, times of day, weather conditions], which may not represent the full range of natural behaviors.
• Observer effect: The presence of drones may influence animal behavior, potentially biasing observations.
• Technical limitations: Drone battery life limited observation sessions to [duration], and weather conditions restricted operations to [conditions].
• Detection bias: Animals may be more difficult to detect in certain environments or weather conditions.

Recommendations

• Users should account for potential observer effects when analyzing behavior patterns.
• Correlations between drone positioning and animal behavior should consider the confounding variables documented in the dataset.
• For machine learning applications, stratified sampling is recommended to address class imbalances in behavior categories.
• When using this data for conservation purposes, consider the ethical implications of drone-based wildlife monitoring.

Licensing Information

This dataset is dedicated to the public domain under the CC0 Public Domain Waiver. We ask that you cite the dataset using the below citation if you make use of it in your research.

Citation

BibTeX:

@misc{kabr_dataset,
  author = {Kline, Jenna M. and Kholiavchenko, Maksim and Brookes, Otto and Berger-Wolf, Tanya and Stewart, Charles V. and Stewart, Christopher},
  title = {KABR Drone Wildlife Monitoring Dataset},
  year = {2024},
  url = {https://kabrdata.xyz/},
  publisher = {Imageomics}
}

Associated Paper:

@misc{kline2024integratingbiologicaldataautonomous,
      title={Integrating Biological Data into Autonomous Remote Sensing Systems for In Situ Imageomics: A Case Study for Kenyan Animal Behavior Sensing with Unmanned Aerial Vehicles (UAVs)}, 
      author={Jenna M. Kline and Maksim Kholiavchenko and Otto Brookes and Tanya Berger-Wolf and Charles V. Stewart and Christopher Stewart},
      year={2024},
      eprint={2407.16864},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2407.16864}, 
}

Acknowledgements

This work was supported by the US National Science Foundation through the Imageomics Institute, which is funded by the Harnessing the Data Revolution (HDR) program under Award #2118240 (Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning). and the AI institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) 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.

Glossary

• KABR: Kenyan Animal Behavior Remote sensing system
• Telemetry: Remote collection of measurement data
• Ethogram: A catalog or inventory of behaviors or actions exhibited by an animal
• Gimbal: A pivoted support that allows rotation of the camera around a single axis
• FPV: First Person View
• Flown state: Operating condition of the drone's flight controller

Dataset Card Authors

Jenna M. Kline, Maksim Kholiavchenko, Otto Brookes, Tanya Berger-Wolf, Charles V. Stewart, Christopher Stewart

Dataset Card Contact

kline dor 377 at osu dot edu