This Paper presents a deep learning approach for recognizing medicinal plants, utilizing a blend of preprocessing techniques and convolutional neural network (CNN) architectures. The proposed method aims to accurately classify a diverse set of medicinal plant species in real-time. The image preprocessing stage employs Canny edge detection, structural derivative analysis, and Gaussian smoothing to enhance the visual feature of the plant leaves. These techniques help to remove noise, sharpen edges, and highlight key morphological characteristics that are crucial for accurate plant identification. The feature extracting stage leverages CNN models, such as ResNet, to automatically learn and extract discriminative features from the preprocessed plant images. The final classification stage employs gradient-based techniques to accurately identify the medicinal plant species. The system is designed to handle a diverse dataset of medicinal plants, certify for the recognition of a widespread of species with high Accuracy. The proposed approach has been extensively evaluated on a comprehensive dataset of medicinal plant images, demonstrating its superior performance compared to traditional manual identification methods. The automated system achieves an accuracy rate of over 97.8%, showcasing its potential for real time deployment in various applications, such as herbal medicine, botanical research, and sustainable plant conservation efforts.