UAV, Handheld
Caltech Aerial RGB-Thermal Dataset
Thermal and RGB-thermal semantic perception and motion tracking
Aerial and on-foot RGB, monochrome, thermal, GPS and inertial recordings of natural US landscapes, with 4195 annotated thermal images and segmentation, translation and motion-tracking benchmarks.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- UAV, Handheld
- Environment
- Outdoor
- Data origin
- Real-world
- Sensor count
- 5
- Ground truth
- Not reported
- Calibration
- Not reported
- Annotations
- Available
Sensor overview
High-level sensor availability before the detailed sensor records below.
Ground truth
Reference scope, method, rate, coverage, and provenance.
Not reported
Calibration and synchronization
Reported calibration level, reproducibility signals, and supporting notes.
- Level
- Not reported
- Processed parameters
- Not reported
- Raw calibration data
- Not reported
- Calibration targets
- Thermal 10x10 circle grid with one-inch circles illuminated by sunlight; RGB and mono 6x6 April tag grid.
Calibration data
Calibration reproducibility details are not reported.
Notes
Kalibr thermal-camera/IMU calibration followed by stereo calibration for each camera pairing; documented calibration procedure does not establish which raw calibration files are released.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
Some recordings omit sensor streams because of collection issues.
Native thermal measurements are not radiometric temperatures.
Rectified RGB-thermal pairs are near-coregistered and have narrower FOV than native thermal images.
Motion-tracking benchmarks use clipped sequences after feature-tracking failures in littoral scenes.
Thermal inversion and geographical appearance changes challenge generalization.
Cameras
thermal
- Model
- Not reported
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Thermal
- Resolution
- 640 x 512
- Rate
- 60 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- 75 deg
- VFOV
- Not reported
- Notes
- Reported capture configuration: non-radiometric FLIR ADK in text, FLIR Boson in Figure 2. Native thermal is 16 bit. Rectified RGB-T pairs have narrower FOV and use the larger RGB image frame. Some recordings lack sensors.
color
- Model
- FLIR Blackfly
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- 960 x 600
- Rate
- 30 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- 75 deg
- VFOV
- Not reported
- Notes
- Footnote prints the shared mono/color designation BFLY-U3-23S6M-C with Tamron M112FM08 lenses. This retains the established family without assigning a guessed per-camera suffix. Capture configuration; not every recording has all streams.
monochrome
- Model
- FLIR Blackfly
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Grayscale
- Resolution
- 960 x 600
- Rate
- 30 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- 75 deg
- VFOV
- Not reported
- Notes
- Companion monochrome camera; footnote shared designation BFLY-U3-23S6M-C with Tamron M112FM08 lenses. Some recordings lack sensors; a per-sequence stream manifest is not supplied by the paper.
IMUs
Sensor-stack IMU
- Model
- VectorNav VN100
- Count
- 1
- Accel / gyro
- Not reported
GNSS
u-blox M8N
- Model or name
- u-blox M8N
- Count
- 1
- Rate
- 5 Hz
- GNSS type
- NOT REPORTED
- Position output
- Available
- Velocity output
- Not reported
- Heading output
- Not reported
LiDAR
Not reported
Radar
Not reported
Pitot tube
Not reported
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Not reported
- Class count
- 10
Semantic segmentation
Available
Instance segmentation
Not reported
Object detection
Not reported
Optical flow
Not reported
Depth ground truth
Not reported
Citation
Citation key and BibTeX kept at the end of the page for reference.
@misc{lee2024cart,
title={Caltech Aerial RGB-Thermal Dataset in the Wild},
author={Lee, Connor and Anderson, Matthew and Ranganathan, Nikhil and Zuo, Xingxing and Do, Kevin and Gkioxari, Georgia and Chung, Soon-Jo},
year={2024},
eprint={2403.08997},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy