ARCHIVES

The study developed an artificial intelligence- based Indian Medicinal Plant Identification and Information System (IMPIS) that makes use of deep learning-based computer vision. In order to perform multi-class classification on 50 species of medicinal plants, the architecture was a two-convolutional neural network (CNN) based on ResNet18 and EfficientNet-B0, which were adjusted using ImageNet weights. The images were pre-processed by uniformly resizing them to 224 by 224 pixels, normalizing them, and image augmentation methods like flipping, zooming, and rotating, with the aim of enhancing the models' generalization and lowering the danger of overfitting. The categorical cross-entropy loss was applied over 50 epochs with early stopping conditions, and Adam was employed as the optimizer with a learning rate of 0.001. Measurements of accuracy, precision, recall, F1-score, and inference latency performance revealed that EfficientNet-B0 could process in real-time (less than 200 ms) with an approximate accuracy of 90.3%. The classification outcome was connected to a knowledge database in the format of a JSON,with a large language model of a JSON using a large language model (Gemini) to produce organized information on phytochemical composition, medicinal use, and administration guidelines. This was created as a complete.