Urban Mistletoe Detection with Multispectral UAV Imagery and Segmentation Masks

1,324 multispectral tiles and 289,956 total images from two UAV flights in Mexico City (September 2022 and May 2023) target the hemiparasitic plant Struthanthus interruptus. The dataset includes preprocessed imagery, derived vegetation indices, texture features, and expert-validated binary segmentation masks. It was created by Paola Andrea Mejia Zuluaga and is associated with a Data Descriptor under review.

Use Cases

• Train semantic segmentation models for mistletoe detection based on the provided binary masks.
• Develop multispectral analysis pipelines using the included vegetation indices and texture descriptors.
• Study temporal changes in urban vegetation based on the two acquisition time points (T1 and T2).
• Benchmark data augmentation techniques using the provided texture (GLCM) and chromatic transformation pipelines.

Strengths

• Includes 1,324 core multispectral tiles and a total of 289,956 images across all data types.
• Provides expert-validated binary segmentation masks for model training and validation.
• Contains preprocessed and derived products including vegetation indices and texture features.
• Data is spatially aligned across multiple modalities (multispectral, RGB, augmentations).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for individual data types is unknown, which may limit suitability assessment.
• Geographic coverage is limited to a single urban park in Mexico City.

Provenance

Source

figshare, author Paola Andrea Mejia Zuluaga.

Collection Method

Captured via UAV flights, with subsequent geometric correction, radiometric correction, and augmentation pipelines.

Time Range

September 2022 and May 2023.

Freshness

Last updated 2026-04-13 18:41:39; freshness should be verified.

Geography

An urban park in Mexico City.