Vegetation Annotation Dataset with 26+ Crop Species

VegAnn contains 3,775 RGB images for semantic segmentation, each 512x512 pixels with corresponding binary masks. The dataset includes images of 26+ crop species captured under diverse outdoor conditions.

Use Cases

• Train a semantic segmentation model to classify vegetation (foreground) from soil and crop residues (background) using the 3,775 image-mask pairs.
• Develop a multi-crop classifier by leveraging the dataset's annotations for 26+ crop species, despite their uneven representation.
• Benchmark segmentation algorithms on agricultural imagery using the provided 512x512 pixel images and their binary masks.

Strengths

• Dataset comprises 3,775 images, providing a substantial volume for model training.
• Images are consistently sized at 512x512 pixels, simplifying preprocessing pipelines.
• Includes annotations for 26+ different crop species, offering diversity in vegetation types.

Limitations

• Crop species are not evenly represented, which may introduce class imbalance for multi-species analysis.
• The binary mask format (0/255) simplifies vegetation/background segmentation but does not distinguish between different crop species.

Provenance

Source

huggingface

Collection Method

Compiled from a variety of outdoor images captured under diverse conditions.

Freshness

Last updated on 2024-02-10.