<![CDATA[Purpose: Purpose: Aristolochia acuminata is a rare plant species of significant conservation value. However, the accurate classification of its parts, such as leaves, stems, and twigs, remains a challenge. This study aimed to develop a reliable classification model to support conservation efforts using Convolutional Neural Network (CNN) technology. Methodology/approach: A digital dataset was systematically collected from various parts of Aristolochia acuminata, forming the foundation for training a CNN-based classification model. To evaluate the model performance and determine the optimal training parameters, three experimental scenarios were conducted using 10, 100, and 200 training epochs. The impact of each training duration on the classification accuracy was analyzed. Results: The model trained with 200 epochs achieved the highest accuracy, outperforming those trained with 10 epochs (68.89%) and 100 epochs (86.67%). This suggests that a longer training period enables the model to learn the visual features of each plant part better, leading to improved classification performance. Conclusion: The results confirm the effectiveness of CNN in classifying the components of Aristolochia acuminata. Using 200 training epochs allowed for deeper feature learning without overfitting, proving optimal in this context. Limitations: This study was limited by the dataset size and the number of classes involved. Further expansion of the dataset and class categories could improve the generalizability of the model. Contribution: This study contributes to plant conservation technology by demonstrating how CNN and structured dataset collection can be applied to classify rare plant species, providing a valuable tool for biodiversity preservation.]]>
Copyrights © 2025
