Evaluating Deep Learning Models for Effective Weed Classification in Agricultural Images

Bianca Panacho Ferreira; Pedro Henrique Campos Moreira; Leandro Henrique Furtado Pinto Silva; João Fernando Mari

doi:10.22456/2175-2745.143490

Authors

Bianca Panacho Ferreira Federal University of Viçosa (UFV) https://orcid.org/0009-0001-5934-0742
Pedro Henrique Campos Moreira Federal University of Viçosa (UFV) https://orcid.org/0009-0005-1871-2676
Leandro Henrique Furtado Pinto Silva Federal University of Viçosa (UFV) https://orcid.org/0000-0002-5765-5206
João Fernando Mari Federal University of Viçosa (UFV) https://orcid.org/0000-0001-8271-3802

DOI:

https://doi.org/10.22456/2175-2745.143490

Keywords:

Deep Learning, Weed Classification, Data Augmentation, CNNs

Abstract

Effective weed management is crucial for maximizing agricultural productivity and minimizing crop losses. Traditional methods for weed detection and classification often suffer from errors and delays, particularly in the absence of specialists. Deep learning technologies offer a promising alternative for automating these tasks, potentially enhancing both accuracy and efficiency. This study compares three advanced deep learning architectures, ResNet-50, EfficientNet V2, and Vision Transformers (ViT), for classifying weed species using the DeepWeeds dataset. We explore the effects of data augmentation on model performance, evaluating each model based on accuracy, precision, recall, and F1 score. Our results demonstrate that data augmentation significantly improves model performance. EfficientNet V2 achieved the highest performance across all metrics, with a peak accuracy of 0.9703. This research provides valuable insights into selecting effective architectures and training strategies for more accurate weed detection in agricultural applications.

Downloads

Download data is not yet available.

References

JR, B. G. M.; CONTINI, E.; ALVES, E. 12. Embrapa: its origins and changes. The regional impact of national policies: the case of Brazil, v. 204, 2012.

ALVES, E. Embrapa: Um caso bem-sucedido de inovação institucional. Revista de Política Agrícola, v. 19, n. 2, p. 65–73, 2010.

Instituto Brasileiro de Geografia e Estatística (IBGE). PIB cresce 1,9% no primeiro trimestre de 2023, impulsionado pela Agropecuária. 2023. https://agenciadenoticias.ibge.gov.br/agencia-noticias/2012-agencia-de-noticias/noticias/37030-pib-cresce-1-9-no-primeiro-trimestre-impulsionado-pela-agropecuaria. Accessed: 2023-12-09.

KUBIAK, A. et al. The problem of weed infestation of agricultural plantations vs. the assumptions of the European biodiversity strategy. Agronomy, v. 12, n. 8, 2022. ISSN 2073-4395. Disponível em: https://doi.org/10.3390/agronomy12081808.

CHAUHAN, B. S. Grand challenges in weed management. Frontiers in Agronomy, v. 1, 2020. Disponível em: https://doi.org/10.3389/fagro.2019.00003.

ADUSEI, F. Y.; ADUSEI, M. A.; LARTEY, B. A roadmap for sustainable disease, pest, and weed management. Biology and Life Sciences Forum, v. 27, n. 1, 2023. ISSN 2673-9976. Disponível em: https://doi.org/10.3390/IECAG2023-14989.

REN, S. et al. Faster R-CNN: Towards real-time object detection with region proposal networks. Advances in neural information processing systems, v. 28, p. 91–99, 2015. Disponível em: https://doi.org/10.1109/TPAMI.2016.2577031.

MUJTABA, G.; BAIK, S. W. Weed detection using deep learning: A systematic literature review. Sensors, MDPI, v. 23, n. 7, p. 3670, 2023. Disponível em: https://doi.org/10.3390/s23073670.

ZHANG, Z. et al. Don’t judge toxic weeds on whether they are native but on their ecological effects. Ecology and Evolution, Wiley Online Library, v. 10, n. 17, p. 9014–9025, 2020. Disponível em: https://doi.org/10.1002/ece3.6609.

OLSEN, A. et al. DeepWeeds: A multiclass weed species image dataset for deep learning. Scientific Reports, Nature Publishing Group UK London, v. 9, n. 1, p. 2058, 2019. Disponível em: https://doi.org/10.1038/s41598-018-38343-3.

SALEEM, M. H. et al. Weed identification by single-stage and two-stage neural networks: A study on the impact of image resizers and weights optimization algorithms. Frontiers in Plant Science, Frontiers Media SA, v. 13, p. 850666, 2022. Disponível em: https://doi.org/10.3389/fpls.2022.850666.

YANG, Y. et al. Dissimilarity-based active learning for embedded weed identification. Turkish Journal of Agriculture and Forestry, v. 46, n. 3, p. 390–401, 2022. Disponível em: https://doi.org/10.55730/1300-011X.3011.

WANG, Y. et al. Fine-grained weed recognition using Swin Transformer and two-stage transfer learning. Frontiers in Plant Science, Frontiers Media SA, v. 14, p. 1134932, 2023. Disponível em: https://doi.org/10.3389/fpls.2023.1134932.

WU, Y.; HE, Y.; WANG, Y. Multi-class weed recognition using hybrid CNN-SVM classifier. Sensors, MDPI, v. 23, n. 16, p. 7153, 2023. Disponível em: https://doi.org/10.3390/s23167153.

HASAN, A. M. et al. Image patch-based deep learning approach for crop and weed recognition. Ecological Informatics, Elsevier, v. 78, p. 102361, 2023. Disponível em: https://doi.org/10.1016/j.ecoinf.2023.102361.

BELISSENT, N. et al. Transfer and zero-shot learning for scalable weed detection and classification in UAV images. Knowledge-Based Systems, v. 292, p. 111586, maio 2024. ISSN 0950-7051.

HE, K. et al. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. [s.n.], 2016. p. 770–778. Disponível em: https://openaccess.thecvf.com/content_cvpr_2016/html/He_Deep_Residual_Learning_CVPR_2016_paper.html.

TAN, M.; LE, Q. EfficientNetV2: Smaller models and faster training. In: PMLR. International Conference on Machine Learning. 2021. p. 10096–10106. Disponível em: https://proceedings.mlr.press/v139/tan21a.html.

DOSOVITSKIY, A. et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. Disponível em: https://doi.org/10.48550/arXiv.2010.11929.

MUMUNI, A.; MUMUNI, F. Data augmentation: A comprehensive survey of modern approaches. Array, v. 16, p. 100258, 2022. ISSN 2590-0056. Disponível em: https://doi.org/10.1016/j.array.2022.100258.

MNIH, V. et al. Human-level control through deep reinforcement learning. Nature, Nature Publishing Group, v. 518, n. 7540, p. 529–533, 2015. Disponível em: https://doi.org/10.1038/nature14236.

LONG, J.; SHELHAMER, E.; DARRELL, T. Fully convolutional networks for semantic segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, p. 3431–3440, 2015. Disponível em: https://doi.org/10.48550/arXiv.1411.4038.