NP-LankaPaddy

The NP-LankaPaddy dataset is an openly accessible, in-field paddy leaf disease image dataset collected under real cultivation conditions in Northern Province of Sri Lanka. It comprises 2,002 RGB images representing four major paddy leaf diseases: brown spot, rice blast, sheath blight, and leaf streak. The images were captured under natural field conditions across four districts in the Northern Province of Sri Lanka: Jaffna, Mannar, Kilinochchi, and Mullaitivu during the two main rice-growing seasons, namely the Maha season (September–March; 1,667 images) and the Yala season (April–August; 335 images). The dataset includes samples from commonly cultivated rice varieties such as Aattakkari, BG406, BG360, and BG300. Image acquisition was performed using mobile phone cameras (iPhone 14 and Samsung Galaxy A54) rather than professional imaging equipment, in order to reflect realistic and practical deployment scenarios. The NP-LankaPaddy dataset exhibits high diversity, encompassing images captured from varied viewing angles, under different lighting conditions, and with diverse backgrounds and environmental variations, without any artificial filtering or post-capture alterations. This diversity ensures that the dataset closely reflects real-world field conditions. Following acquisition, all images were resized to a uniform resolution of 768 × 1024 pixels. The dataset is organised into training (1,400 images) and testing (602 images) subsets.