Comprehensive Herb Species Dataset: Leveraging Deep Learning for Accurate Identification and Classification
Description
1. Herbs hold significant value across various sectors, including traditional medicine, culinary arts, and cosmetics, due to their unique properties and health benefits. However, accurately identifying different herb species can be challenging. The Herb Species Dataset is designed to aid researchers, horticulturists, and machine learning professionals in identifying and categorizing various herb species. This dataset comprises high-resolution images of herb leaves, showcasing both different species and various conditions. It includes several common herb species such as Azadirachta indica (Neem), Centella Asiatica (Gotu Kola), Coriandrum sativum (Coriander), Ocimum sanctum (Holy Basil), and Cinnamomum tamala (Indian Bay Leaf). Extensive annotations are provided for each image, offering detailed information regarding the species and condition of the leaves, thus enabling accurate and dependable model training and validation. Furthermore, the dataset is enhanced with metadata documenting the geographical locations and environmental conditions at the time of image capture. This contextual data can improve the precision of predictive models and is essential for comprehending the factors influencing herb identification. 2. These days, deep learning and computer vision techniques hold significant promise for handling classification and detection tasks, including herb species identification. Leveraging these technologies can significantly improve the accuracy and efficiency of such tasks. 3. A thorough dataset for herb species identification is presented to develop deep learning methods. This dataset's classifications were created in cooperation with a subject matter expert from a botanical institute. The expert's involvement ensures that the classifications are accurate and reliable, providing a solid foundation for model development. 4. A total of 2,500 images of the five herb species were gathered from various locations. To enhance the number of data points, 30,000 augmented images were created from these original images by applying flipping, width shifting, height shifting, brightening, rotating, shearing, and zooming. This extensive augmentation process ensures that the dataset is robust and capable of supporting the development of highly accurate predictive models.