Sign in to view source links and access this dataset
Description
BDD100K-validation contains 10,000 samples from one of the largest open-source driving datasets. The dataset consists of images extracted every 10 seconds from driving videos, providing labels for object detection, weather, time of day, and scene. It was uploaded to Hugging Face by dgural and last updated on June 12, 2024.
Use Cases
Train object detection models based on labeled bounding boxes mentioned in the description.
Analyze driving scene context based on provided weather and time-of-day labels.
Develop models for robust perception across diverse environmental conditions based on the varied scene labels.
Benchmark autonomous vehicle perception systems on a large-scale, open-source dataset.
Strengths
Contains 10,000 samples in this validation split.
Provides multiple annotation types including object detection, weather, time of day, and scene labels.
Derived from one of the largest open-source driving datasets.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count for the full dataset is unknown, which may limit suitability assessment.
Freshness should be verified as the last update was 2024-06-12.
Provenance
Source
BDD100K dataset, uploaded to Hugging Face by dgural.
Collection Method
Images extracted every 10 seconds from driving videos.
Freshness
Last updated 2024-06-12.
Requires installation of the FiftyOne library (`pip install -U fiftyone`) for usage as indicated in the description.