BanSpot: A Field-Acquired Banana Leaf Image Dataset for Leaf Spot Symptom Identification

BanSpot (Banana Leaf Spot Symptom dataset) is a field-acquired banana leaf image dataset containing leaf spot symptoms and non leaf spot banana leaf samples collected under natural agricultural conditions from 10 farms distributed across various locations in Kasaragod district, Kerala, India.

This dataset contains banana leaf images exhibiting visible leaf spot symptoms as well as visually healthy or asymptomatic leaves captured under real-world field environments using smartphone cameras. Images were collected under varying environmental conditions including differences in lighting, viewing angle, background complexity, and symptom severity.

Under natural agricultural field conditions, visually similar foliar spotting patterns may arise from multiple biotic factors, making strict disease-specific annotation challenging without laboratory confirmation. Therefore, the BanSpot dataset adopts a symptom-oriented annotation strategy focused on the presence or absence of leaf spot symptoms under real-world field environments.

Additionally, non leaf spot artifacts such as insect bite marks, dust and debris in the surface of leaves were intentionally retained within the both symptomatic and asymptomatic classes to preserve realistic field conditions. This strategy was adopted to minimize the chances of future models associated such artifacts as strictly symptomatic or asymptomatic, thereby improving model robustness.

Files included:

DATASET.zip, METADATA.xlsx, README.txt and Specifications table.docx

This dataset is intended for research and educational purposes in plant disease recognition, computer vision, deep learning, and agricultural artificial intelligence.