Sign in to view source links and access this dataset
Description
Robotflowlabs created a collection of 1,718,541 drone images, each with a caption. The dataset includes images across six modalities: original RGB, night, dusk, fog, rain, and thermal. It was last updated on April 16, 2026.
Use Cases
Train object detection models based on the 1.4 million YOLO labels mentioned in the description.
Develop and test image captioning models for aerial imagery using the provided captions.
Benchmark model robustness across different environmental conditions based on the five synthesized adverse conditions (night, dusk, fog, rain, thermal).
Train domain translation models using the five trained translation models referenced in the description.
Strengths
Contains 1,718,541 captioned images, providing a large-scale resource.
Covers six distinct imaging modalities, including five synthesized adverse conditions.
Includes 1.4 million YOLO labels for object detection tasks.
Provides a reproducible pipeline for generating the dataset.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
robotflowlabs
Collection Method
Likely gathered and synthesized from a single drone platform, with conditions artificially generated.
Freshness
Last updated 2026-04-16 10:43:29; freshness should be verified.
License restrictions are unknown and should be verified before use.