DroneScapes2: 218 Annotated Drone Images for Segmentation and Classification

DroneScapes2 Annotated Train Set provides 218 drone-view images with segmentation and classification labels, curated by Voxel51. This dataset is a specific subset of the larger DroneScapes2 repository, formatted for use with the FiftyOne library as of late 2025.

Use Cases

• Training semantic segmentation models using the provided image masks
• Fine-tuning image classification algorithms on aerial drone perspectives
• Visualizing and auditing drone-based computer vision labels within the FiftyOne tool

Strengths

• 218 annotated samples
• Dual-task support for classification and segmentation
• Native integration with the FiftyOne library for visualization

Limitations

• Small sample size of 218 records
• Geographic bias restricted to the United States region
• Lack of detailed column metadata or class distribution in the source description

Provenance

Source

Voxel51

Collection Method

Annotated subset from the original DroneScapes2 repository

Freshness

Last updated November 2025.

Geography

United States