Enhancing Succulent Plant Species Classification: A MobileNetV3-Large Approach

Recent advancements in deep learning have significantly enhanced the classification accuracy of various plant species, including succulents. This study aims to evaluate the performance of the MobileNetV3-Large architecture in classifying 10 types of succulent plants, comparing its accuracy to previous models, and identifying potential areas for improvement. A quantitative study was conducted using the MobileNetV3-Large architecture, with hyperparameters set to input shape $224 \times 224 \times 3$ (RGB channel), batch size 64, optimizer Adam, learning rate 0.0001, and 30 epochs. The model was trained and validated on a dataset of 4,500 images, with performance metrics including accuracy and loss values for both training and validation data. The model achieved an accuracy of $\mathbf{99.11\%}$ across 4,500 samples. Specific species such as Agave Parryi, Crassula Ovata, and others reached 100% accuracy for each sample tested. Minor inaccuracies were observed in Kalanchoe Luciae and Sempervivum Tectorum, with one sample each showing slightly lower accuracy (99.95% and 93.59%, respectively). Compared to previous studies, this model maintained high performance with a larger dataset. The results demonstrate that the MobileNetV3-Large model performs exceptionally well in classifying succulent plant species, achieving higher accuracy with a larger dataset compared to previous models. While the model’s performance is robust, minor misclassifications indicate areas for future improvement. Further research should focus on optimizing hyperparameters and increasing dataset diversity to enhance model accuracy and generalizability