Automotive
BDD100K
Multitask driving perception and tracking
Crowdsourced driving clips with task-specific image, lane, drivable-area, semantic, instance and tracking labels.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Automotive
- Environment
- Outdoor
- Data origin
- Real-world
- Sensor count
- 3
- 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
- Not reported
Calibration data
Calibration reproducibility details are not reported.
Notes
Not reportedKnown or intentional limitations
Reported constraints and characteristics to check before using the dataset.
The 10K segmentation-image package overlaps with but is not strictly a subset of the 100K keyframes/videos. Detection, semantic, panoptic, tracking and later pose products have separate membership and annotation coverage.
The 100K forty-second clips give a nominal four million seconds, not independently measured unique ride time. Tracking labels/images are resampled at 5 Hz from 30 fps video.
Maintained documentation recommends revised 2020 detection labels. Object, semantic, panoptic and tracking taxonomies have different evaluated classes; no single aggregate class count is assigned.
Crowdsourced camera/navigation models, physical counts, native IMU/GNSS rates, calibration and independent reference accuracy remain unknown. GPS/IMU recordings are not independent pose ground truth.
Official GitHub documentation identifies the ETH data destination. Main-site requests timed out, documentation DNS failed and direct archive indexing returned 502 during review; current bulk reachability is unverified.
Cameras
Camera 1
- Model
- Not reported
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- Not reported
- Rate
- 30 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Crowdsourced videos; physical camera model/count and exact width are not established by 720p.
IMUs
Crowdsourced video-associated IMU
- Model
- Not reported
- Count
- 1
- Accel / gyro
- Not reported
GNSS
GNSS 1
- Model or name
- Not reported
- Count
- 1
- Rate
- Not reported
- GNSS type
- UNKNOWN
- Position output
- Available
- Velocity output
- Not reported
- Heading output
- Not reported
LiDAR
Not reported
Radar
Not reported
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Task-specific Scalabel JSON, masks/bitmasks, lane/drivable-region attributes and temporal tracks; later panoptic and pose_21 products are documented separately. Native and evaluated taxonomies vary by task.
- Class count
- Not reported
Semantic segmentation
Available
Instance segmentation
Available
Object detection
Available
Optical flow
Not reported
Depth ground truth
Not reported
Citation
Citation key and BibTeX kept at the end of the page for reference.
@inproceedings{yu2020bdd100k,
author = {{Fisher Yu} and {Haofeng Chen} and {Xin Wang} and {Wenqi Xian} and {Yingying Chen} and {Fangchen Liu} and {Vashisht Madhavan} and {Trevor Darrell}},
title = {BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning},
year = {2020},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition},
eprint = {1805.04687},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/1805.04687v2}
}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy