Abstract
Medicinal plants are a potential source of income particularly for rural populations in India who rely on these medicinal plants to treat a variety of different diseases through both targeted temporary and daily use. Through this data paper we have given a reference to our collected specimen set where almost 117 (one hundred seventeen) medicinal plant species with their leaf samples are stored. We have used Mendeley platform to store the dataset and visited many medicinal plant gardens situated in Assam to collect them. The dataset consists of raw leaf samples, U-net segmented gray leaf samples and a plant name table. The table includes Botanical name, family of the species, Common name and Assamese name. For segmentation U-net model was used and the resultant U-net segmented gray image frames are uploaded to the database. These segmented samples can be directly used for deep learning model for training and classification. Researchers will be able to use those to build recognition tool for Android or PC based system.