ExDARK: Low-Light Images with 10 Lighting Conditions and Object Annotations

Featuring 10 distinct low-light conditions, this dataset provides images with both image-level class and object-level annotations for computer vision research. Created by cs-chan and updated in 2026, it targets tasks like object detection and image translation in environments ranging from twilight to near-total darkness.

Use Cases

• Object detection in dark environments using the object-level annotations
• Image classification using the 10 different low-light condition labels
• Low-light image enhancement using the twilight and very low-light image categories

Strengths

• Includes 10 distinct low-light condition categories
• Provides both image-level and object-level annotations
• Licensed under the BSD-3-Clause for open redistribution

Limitations

• Total record count is not specified in the source metadata
• Specific object classes are not enumerated in the provided description

Provenance

Source

cs-chan via GitHub

Collection Method

annotated

Freshness

Last updated February 2026.