Skip to content

Dataset · dgural

BDD100k road scenes validation split

dgural/bdd100k

10,000 dashcam frames with object, weather, and time-of-day annotations for autonomous vehicles.

What's inside

A split from the well-known open autonomous driving dataset BDD100k, prepared for use in the FiftyOne library. It contains 10,000 images (every 10th second of video recordings) with English-language annotations. The annotations specify bounding boxes for object detection, weather conditions, time of day, and road scene type.

What you can do with it

  • Validating object detection and road scene classification models
  • Evaluating the robustness of autonomous vehicle computer vision to varying weather and lighting
  • Rapid prototyping and visual error analysis in FiftyOne
Tasks
Image classification, Object detection
Languages
English
Size
10K–100K rows
Downloads this month
22.6K
Likes
5

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("dgural/bdd100k")