Deep Herb: Ayurveda Dry Herb Recognition and Question Answering Using Deep Learning

Dry herbs play a pivotal role in the daily lives of common people, offering a diverse array of health benefits and culinary delights. From enhancing the flavor and aroma of dishes to serving as natural remedies for various ailments, dry herbs are indispensable in kitchens and medicine cabinets alike. Identification and recognition of such indigenous herbs becomes very vital for its popularization among common people. With this objective, this research work aims to build deep learning models “Deep Herb” to recognize dry herbs that are used in traditional Indian medicine and cuisine using a simple mobile application.. Three deep learning models namely Convolutional Neural Network (CNN), standard and modified VGG19 models are trained on a comprehensive dataset of dry herb images, encompassing diverse species and variations. Data augmentation techniques are used to enhance dataset diversity, while preprocessing methods such as resizing and noise reduction are utilized for optimizing the image quality. During the training phase, careful adjustments are made using transfer learning to fine-tune the model's parameters aiming to achieve optimal accuracy and robustness. Notably, the modified VGG19 model achieves an impressive accuracy rate of98%. Finally the Deep Herb model is deployed in an Android mobile application through which users can interact with the model. Using this Deep Herb mobile application users can take pictures of any dry herb, submit the image of the herb to the model and recognise them with ease.