Castor Beans Identification in Sugarcane Plantations Using Active Learning

Juan Carlos Conceiç˜ao de Lima Sales; Mário de Araújo Carvalho; José Marcato Junior; Wesley Nunes Goncalves

doi:10.22456/2175-2745.143495

Authors

Juan Carlos Conceiç˜ao de Lima Sales Universidade Federal de Mato Grosso do Sul (UFMS) https://orcid.org/0000-0002-7590-6275
Mário de Araújo Carvalho Universidade Federal de Mato Grosso do Sul (UFMS)
José Marcato Junior Universidade Federal de Mato Grosso do Sul (UFMS)
Wesley Nunes Goncalves Universidade Federal de Mato Grosso do Sul (UFMS)

DOI:

https://doi.org/10.22456/2175-2745.143495

Keywords:

Active Learning, Machine Learning, Weed Control, Castor Beans

Abstract

Castor beans have multiple applications but can become weeds in several crops. Manual identification is unfeasible in large plantations. Machine learning models can make identification efficient but require a lot of labeled data. Many images of a plantation would be similar and wouldn't improve the model's performance. Active learning (AL) allows labeling only relevant data, often surpassing the performance of models trained on entire datasets. In this work, we detect castor beans, with an AL method that uses self-supervised pretext tasks to separate data for labeling. Models based on pretext tasks presented a decrease in recall relative to the model trained in the whole dataset, which has 93%. We also trained a pseudo-task that separates data with a reasonable concentration of castor bean images. The pseudo-task classifier obtained a 92% recall, being trained in less than 1% of the dataset.

Downloads

Download data is not yet available.

References

IN: CONGRESSO BRASILEIRO DE MAMONA, 3., 2008, SALVADOR. ENERGIA E... Utilização de coprodutos da mamona na alimentação animal. Disponível em: ⟨https://www.alice.cnptia.embrapa.br/alice/handle/doc/939880⟩.

SAADAOUI, E. et al. Castor bean (Ricinus communis L.) diversity, seed oil and uses. Oilseed Crops: Yield and Adaptations under Environmental Stress, Wiley Online Library, p. 19–33, 2017. Disponível em: ⟨https://doi.org/10.1002/9781119048800.ch2⟩.

LIMA, R. et al. Casca e torta de mamona avaliados em vasos como fertilizantes orgânicos. Revista Caatinga, v. 21, n. 5, p. 102-106, supl., dez., 2008. Disponível em: ⟨https://www.alice.cnptia.embrapa.br/alice/handle/doc/277960⟩.

ZERA, F. S. et al. Tolerância de mamona (Ricinus communis) a herbicidas utilizados na cultura da cana-de-açúcar. Nucleus, Fundação Educacional Ituverava, v. 8, n. 1, p. 1–10, 2011. Disponível em: ⟨https://doi.org/10.3738/1982.2278.469⟩.

FONSECA, N. B. da S.; SOTO-BLANCO, B. Toxicidade da ricina presente nas sementes de mamona. Semina: Ciências Agrárias, Universidade Estadual de Londrina, v. 35, n. 3, p. 1415–1424, 2014. Disponível em: ⟨https://doi.org/10.5433/1679-0359.2014v35n3p1415⟩.

OLIVEIRA, R. C. de; FELIZARI, J. Toxicidade de extratos de Ricinus communis e Euphorbia milii para caramujo aquático e tilápia. Revista Thêma et Scientia, v. 1, n. 2, 2011. Disponível em: ⟨https://ojsrevistas.fag.edu.br/index.php/RTES/article/view/467⟩.

COSTA, N. V. et al. Métodos de controle de plantas daninhas em sistemas orgânicos: breve revisão. Revista Brasileira de Herbicidas, v. 17, n. 1, p. 25–44, 2018. Disponível em: ⟨https://doi.org/10.7824/rbh.v17i1.522⟩.

FERREIRA, A. d. S. Redes neurais convolucionais profundas na detecção de plantas daninhas em lavoura de soja. 2017. Disponível em: ⟨https://repositorio.ufms.br/handle/123456789/3101⟩.

ODA, Y. S.; NETO, J. C. d. C. Detecção de plantas daninhas e plantas de soja utilizando imagens multiespectrais e visão computacional. 2023. Disponível em: ⟨https://doi.org/10.11606/T.76.2023.tde-22022024-084207⟩.

PHAM, T.-N. et al. Improved YOLOv5 based deep learning system for jellyfish detection. IEEE Access, IEEE, 2024. Disponível em: ⟨https://doi.org/10.1109/ACCESS.2024.3405452⟩.

SETTLES, B. Active learning literature survey. University of Wisconsin-Madison Department of Computer Sciences, 2009. Disponível em: ⟨http://digital.library.wisc.edu/1793/60660⟩.

YI, J. S. K. et al. Using self-supervised pretext tasks for active learning. CoRR, abs/2201.07459, 2022. Disponível em: ⟨https://arxiv.org/abs/2201.07459⟩.

MOLINA-VILLA, M. A.; SOLAQUE-GUZMÁN, L. E. et al. Machine vision system for weed detection using image filtering in vegetables crops. Revista Facultad de Ingeniería Universidad de Antioquia, Universidad de Antioquia, n. 80, p. 124–130, 2016. Disponível em: ⟨https://doi.org/10.17533/udea.redin.n80a13⟩.

PEREZ, A. et al. Colour and shape analysis techniques for weed detection in cereal fields. Computers and Electronics in Agriculture, Elsevier, v. 25, n. 3, p. 197–212, 2000. Disponível em: ⟨https://doi.org/10.1016/S0168-1699(99)00068-X⟩.

LIN, F. et al. Detection of corn and weed species by the combination of spectral, shape and textural features. Sustainability, MDPI, v. 9, n. 8, p. 1335, 2017. Disponível em: ⟨https://doi.org/10.3390/su9081335⟩.

ANDÚJAR, D.; WEIS, M.; GERHARDS, R. An ultrasonic system for weed detection in cereal crops. Sensors, Molecular Diversity Preservation International (MDPI), v. 12, n. 12, p. 17343–17357, 2012. Disponível em: ⟨https://doi.org/10.3390/s121217343⟩.

SHAHBAZI, N. et al. Assessing the capability and potential of lidar for weed detection. Sensors, MDPI, v. 21, n. 7, p. 2328, 2021. Disponível em: ⟨https://doi.org/10.3390/s21072328⟩.

PUNITHAVATHI, R. et al. Computer vision and deep learning-enabled weed detection model for precision agriculture. Comput. Syst. Sci. Eng., v. 44, n. 3, p. 2759–2774, 2023. Disponível em: ⟨https://doi.org/10.32604/csse.2023.027647⟩.

IEEE. Deep Learning Based Weed Detection and Elimination in Agriculture. p. 147–151. Disponível em: ⟨https://doi.org/10.1109/ICICT57646.2023.10134186⟩.

Active learning approaches to enhancing neural machine translation. p. 1796–1806. Disponível em: ⟨https://doi.org/10.18653/v1/2020.findings-emnlp.162⟩.

IEEE. Modulation and signal class labelling with active learning and classification using machine learning. p. 1–5. Disponível em: ⟨https://doi.org/10.1109/CONECCT55679.2022.9865826⟩.

FERREIRA, M. D. et al. Active learning for medical code assignment. 2021. Disponível em: ⟨https://arxiv.org/abs/2104.05741⟩.

IEEE. Active machine learning to test autonomous driving. p. 286. Disponível em: ⟨https://doi.org/10.1109/ICSTW52544.2021.00055⟩.

CARVALHO, M. de A. DalMax: Framework for Deep Active Learning Approaches. 2024. ⟨https://github.com/MarioCarvalhoBr/dalmax-framework-deep-active-learning-python⟩. Available on GitHub.

LEWIS, D. D.; GALE, W. A. A sequential algorithm for training text classifiers. CoRR, abs/cmp-lg/9407020, 1994. Disponível em: ⟨http://arxiv.org/abs/cmp-lg/9407020⟩.