The unique approach discussed here employs CNNs and deep learning in a vision-based system to automatically detect plants. This research aims to improve medicinal plant identification and categorisation using CNNs and pixel-based evaluations. This research uses Convolutional Neural Networks (CNNs) to analyse medicinal plant photos at various resolutions (64 × 64 to 256 × 256 pixels). Through pixel-level analysis, plant images may be examined, and delicate visual traits extracted for precise identification. It is important to consider LSTM in recurrent neural networks. It helps readers grasp sequential data patterns, making it ideal for image data analysis that requires context. To accurately classify plants, the system's deep learning algorithms independently acquire and discriminate these characteristics. Convolutional Neural Networks (CNNs) and pixel-based assessments are used in the “Herbal Detector” project to recognise five medicinal plants with precision above 99.6%. The system's accuracy across resolutions shows its endurance and versatility.