RiceSeg-5932: Complete Pixel-Level Segmentation Masks for Rice Leaf Disease Images Samples

This dataset provides expert-annotated, pixel-level segmentation masks for the publicly available collection of 5,932 rice leaf images originally released by Sethy et al. (2020). The images cover four major rice leaf diseases: Bacterial Blight, Leaf Blast, Brown Spot, and Tungro, representing diverse disease manifestations.

Each image has a corresponding manual segmentation mask in JPG format, created using "Label Studio" to precisely delineate diseased regions at the pixel level. These ground-truth masks enable supervised training, evaluation, and benchmarking of semantic segmentation models for plant disease detection, supporting reproducible research in agricultural computer vision.

This dataset represents the first publicly available, complete set of pixel-level segmentation masks for all images in the original collection, providing a valuable resource for machine learning, computer vision research, and quantitative disease-region analysis in rice crops.

❗ IMPORTANT: How to Use This Dataset

This dataset contains only the segmentation masks. To use it correctly:

- Download the original images from:
sethy, prabira Kumar (2020), “Rice Leaf Disease Image Samples”, Mendeley Data, V1, doi: 10.17632/fwcj7stb8r.1
- Pair masks and images using exact filename correspondence with Version 1 or Vesion 2 of the dataset.