📄 Sciences Methods and Technologies
International Journal (SciMeTech)

Volume 2 · Issue 2 · 2026
ISSN: 3085-5284
Lightweight Damascena Rose Maturity Classification via Wavelet Scattering
Mohamed Ohamouddou, Rafik Lasri, Suhail Odeh
Pages 197–202 · 1. TED-AEEP Research Group, FPL, Abdelmalek Essaadi University, Tetouan, Morocco · 2. Software Engineering Department, ASTE Faculty, Bethlehem University, Bethlehem P125, Palestine
⬇ PDF
Abstract
Rosa damascena, commonly known as the Damascus rose, is one of the most widely cultivated rose species for natural essential oil and rose water. In Morocco, its cultivation is concentrated in the Dades Valley around Kelaa M'Gouna, where the crop has long been a defining part of the local economy and where an annual Rose Festival is held at the end of the harvest. The harvest is short and time-sensitive: petals must be picked at a specific stage of bud opening and before the morning sun warms the field, because the volatile aromatic compounds evaporate quickly as the temperature rises. Sorting is done by hand, and pickers judge each bud visually under variable light. An automated visual classifier on a hand-held device could improve consistency and reduce labour cost, but only if the underlying model is small enough for the hardware available in the field.

Convolutional neural networks (CNNs) are now a standard tool in agricultural image analysis. Most systems start from a large ImageNet pretrained backbone such as VGG or ResNet and fine-tune it on the target dataset. The recipe transfers to rose maturity, but the resulting models are heavy, well beyond the budget of low-power devices. To address this, several families of compact CNNs have been developed. However, a complementary line of work asks whether the early visual features need to be learned at all. The wavelet scattering transform cascades wavelet convolutions with a modulus non-linearity to produce a representation that is locally translation-invariant and Lipschitz stable to small deformations.

In agricultural settings, where datasets are typically small and on-device deployment is required, the inductive bias of fixed wavelet features can compensate for limited supervision while keeping the gradient and parameter footprint small. To the best of our knowledge, this design has not yet been evaluated on a floriculture task. The contributions of this paper are: (i) a binary rose maturity classifier that uses a 2D Morlet wavelet scattering transform (J = 3, L = 8) as a frozen front-end followed by a 61,410-parameter convolutional head; (ii) a benchmark against six ImageNet-pretrained baselines (SqueezeNet 1.1, MobileNetV3-Small, ShuffleNetV2-x0.5, MobileNetV2, EfficientNet-B0, and ResNet50) trained under an identical protocol; and (iii) a trade-off analysis showing 96.0% accuracy and an F1-score of 0.96 with 5.6 to 383 times fewer trainable parameters than the baselines, at the cost of a 2.2 percentage-point accuracy gap.
Keywords: Rosa damascena, Rose harvesting, Wavelet scattering, Precision agriculture, Lightweight neural network

References

  1. Mallat, S. (2012). Group invariant scattering. Communications on Pure and Applied Mathematics, 65(10), 1331-1398.
  2. Bruna, J., & Mallat, S. (2013). Invariant scattering convolution networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8), 1872-1886.
  3. Andreux, M., Angles, T., Exarchakis, G., Leonarduzzi, R., Rochette, G., Thiry, L., Zarka, J., Mallat, S., Anden, J., Belilovsky, E., Bruna, J., Lostanlen, V., Chaudhary, M., Hirn, M. J., Oyallon, E., Zhang, S., Cella, C., & Eickenberg, M. (2020). Kymatio: Scattering transforms in Python. Journal of Machine Learning Research, 21(60), 1-6.
  4. Oyallon, E., & Mallat, S. (2015). Deep roto-translation scattering for object classification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2865-2873.
  5. Oyallon, E., Belilovsky, E., Zagoruyko, S., & Valko, M. (2019). Scattering networks for hybrid representation learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(9), 2208-2221.
  6. Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016). SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and < 0.5 MB model size. arXiv preprint arXiv:1602.07360.
  7. Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017). MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
  8. Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 4510-4520.
  9. Howard, A., Sandler, M., Chu, G., Chen, L.-C., Chen, B., Tan, M., Wang, W., Zhu, Y., Pang, R., Vasudevan, V., Le, Q. V., & Adam, H. (2019). Searching for MobileNetV3. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 1314-1324.
  10. Ma, N., Zhang, X., Zheng, H.-T., & Sun, J. (2018). ShuffleNet V2: Practical guidelines for efficient CNN architecture design. In Proceedings of the European Conference on Computer Vision (ECCV), 116-131.
  11. Tan, M., & Le, Q. (2019). EfficientNet: Rethinking model scaling for convolutional neural networks. In Proceedings of the 36th International Conference on Machine Learning (ICML), Vol. 97, 6105-6114.
  12. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770-778.
  13. Mohanty, S. P., Hughes, D. P., & Salathe, M. (2016). Using deep learning for image-based plant disease detection. Frontiers in Plant Science, 7, 1419.
  14. Kamilaris, A., & Prenafeta-Boldu, F. X. (2018). Deep learning in agriculture: A survey. Computers and Electronics in Agriculture, 147, 70-90.
  15. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. In Advances in Neural Information Processing Systems (NeurIPS), 25, 1097-1105.
  16. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1-9.
  17. Simonyan, K., & Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. In International Conference on Learning Representations (ICLR).
  18. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). ImageNet: A large-scale hierarchical image database. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 248-255.
  19. Liakos, K. G., Busato, P., Moshou, D., Pearson, S., & Bochtis, D. (2018). Machine learning in agriculture: A review. Sensors, 18(8), 2674.
  20. Ohamouddou, M., Ohamouddou, S., El Afia, A., & Lasri, R. (2025). ATMS-KD: Adaptive temperature and mixed sample knowledge distillation for a lightweight residual CNN in agricultural embedded systems. Smart Agricultural Technology, 12, 101617.
SciMeTech · Volume 2, Issue 2 · 2026 · ISSN 3085-5284