Automatic Cactus Species Identification using Deep Learning Model and Object Detection Techniques
Keywords:
image classification, cactus species, deep learning, object detectionAbstract
Identifying cactus species can be challenging for beginners due to the diversity of species and the similarities in distinctive features, which may affect the accuracy of classification. This research aims to 1) develop a dataset of images and image descriptions for cactus species, 2) develop a cactus species identification model using deep learning and object detection techniques, and 3) evaluate the accuracy of cactus species identification. The work involves collecting images of 10 popular cactus species from commercial cactus growers in Nakhon Ratchasima province, along with creating image descriptions in XML format, totaling 1,508 images. The MobileNetV2 model, which has been pre-trained with transfer learning, was used to train on the training dataset. The optimizer used was RMSprop, with an appropriate loss function to improve model accuracy during backpropagation for adjusting the model weights to suit the key features of each cactus species. Additionally, a single-shot object detection technique was employed for fast and efficient object detection and classification. The results from training the model showed an average accuracy of 92.64% on the training dataset and 91.05% on the validation dataset, indicating that the model was able to effectively learn data patterns without significant signs of overfitting or underfitting. The average precision and average recall for large object detection were high, at 0.991 and 0.993, respectively. However, limitations in detecting medium and small objects were observed and require future improvement. When evaluating the model in real-world scenarios with unseen datasets, the average accuracy was 83.33%. The model performed well in cases where a single object could be classified into multiple classes or when multiple objects were in the same class. However, accuracy decreased when images contained multiple objects from different classes.
References
Ali, M., Keller, C., & Huang, M. (2023). Fruits detection using single shot multibox detector. In BSCI ‘23: Proceedings of the 5th ACM International Symposium on Blockchain and Secure Critical Infrastructure (pp. 140–144). New York: Association for Computing Machinery
Anderson, E. F. (2001). The cactus family. Portland, OR: Timber Press.
Ariwa, R. N., Markus, C., Teneke, N. G., Adamu, S., & George, F. K. (2024). Plant disease detection using YOLO machine learning approach. British Journal of Computer, Networking and Information Technology, 7(2), 115–129. https://doi.org/10.52589/BJCNIT-EJWGFW6D
Arun, A., Sharma, S., Singh, B., & Hazra, T. (2025). Identification of plant species using convolutional neural network with transfer learning. Journal of Phytopathology, 173(1). e70032. https://doi.org/10.1111/jph.70032
Barthlott, W., & Hunt, D. R. (1993). Cacti: An illustrated guide to the plants and their cultivation. Portland, OR: Timber Press.
Business Research Insights. (2023). Succulent plant market report (Market Report No. 100171). Retrieved from https://www.businessresearchinsights.com/market-reports/succulent-plant-market-100171
Girshick, R. (2015). Fast R-CNN. In Proceedings of the IEEE International Conference on Computer Vision (ICCV) (pp. 1440–1448). Santiago: IEEE
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. Cambridge, MA: MIT Press.
Gong, X., & Zhang, S. (2023). An analysis of plant diseases identification based on deep learning methods. The Plant Pathology Journal, 39(4), 320–324. https://doi.org/10.5423/PPJ.OA.02.2023.0034
Kaur, S., & Kaur, P. (2019). Plant species identification based on plant leaf using computer vision and machine learning techniques. Journal of Multimedia Information System, 6(2), 49–59. https://doi.org/10.33851/JMIS.2019.6.2.49
Konyushkova, K., Pishchulin, L., & Matuszek, C. (2018). Learning intelligent dialogs. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 1–10). UT: IEEE
Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.-Y., & Berg, A. C. (2016). SSD: Single shot multibox detector. In B. Leibe et al. (Eds.), Proceedings of the European Conference on Computer Vision (ECCV 2016) (pp. 21–37). Amsterdam: Springer International Publishing.
Nobel, P. S. (2002). Cacti: Biology and uses. CA: University of California Press.
Rani, A., Nikhila, M., Bhargavi, M., & Sushma, P. (2024). Supervised learning based plant species classification for precise e-agriculture. Journal for Educators, Teachers and Trainers, 15(5), 405–415. Retrieved from https://hdl.handle.net/10481/103749
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 779–788). NV: IEEE
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. arXiv, 1801(04381), 4510-4520. https://doi.org/10.48550/arXiv.1801.04381
Thomkaew, J., & Intakosum, S. (2023). Plant species classification using leaf edge feature combination with morphological transformations and SIFT key point. Journal of Image and Graphics, 11(1), 91–97. https://doi.org/10.18178/joig.11.1.91-97
Verified Market Research. (2021). Cactus extract market – Global industry trends and forecast to 2027. Retrieved from https://www.verifiedmarketresearch.com/product/cactus-extract-market/
Wu, H., Fang, L., Yu, Q., & Yang, C. (2024). Composite descriptor based on contour and appearance for plant species identification. Engineering Applications of Artificial Intelligence, 133(3), 108291. https://doi.org/10.1016/j.engappai.2024.108291
Zhang, X., Yang, Y.-H., Han, Z., Wang, H., & Gao, C. (2014). Object class detection: A survey. ACM Computing Surveys, 46(1), Article 10, 1–53. https://doi.org/10.1145/2522968.2522978
