Advanced Plant Species Classification with Interclass Similarity using a Fine-Grained Computer Vision Approach

Plant species classification holds immense importance across diverse sectors such as agriculture, environmental conservation, and botanical research. Accurate identification of plant species is fundamental for effective ecological management, agricultural practices, and biodiversity preservation. In this context, the research addresses the critical task of fine-grained plant classification, specifically emphasizing vegetable and fruit plants. Recognizing the significance of visual similarity analysis in plant sciences, a robust methodology leveraging Convolutional Neural Networks (CNN) is introduced. The hybrid CNN model combines the interpretability of VGG16 with the gradient-stabilizing benefits of ResNet50's residual connections, enabling nuanced classification of 23 plant classes. Achieving an outstanding 99% accuracy in plant class prediction, the model is not only highly accurate but also interpretable. Additionally, the incorporation of cosine similarity enhances the robustness of the approach, facilitating a leaf similarity search application. This innovative feature strengthens the identification of visually similar plant images based on extracted features, significantly contributing to the advancement of plant species classification methodologies.