Pasar al contenido principal

AVANCES EN DETECCION PROXIMA DE MALAS HIERBAS Y APLICACION DE TRATAMIENTOS MIXTOS PARA REDUCIR EL USO DE HERBICIDAS Y FOMENTAR LA AGRICULTURA ECOLOGICA

PID2023-150108OB-C33

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para Impulsar la Investigación Científico-Técnica y su Transferencia
Subprograma Subprograma Estatal de Generación de Conocimiento
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2023
Plan Estatal 2021-2023 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario AGENCIA ESTATAL CONSEJO SUPERIOR DE INVESTIGACIONES CIENTIFICAS (CSIC)

Resultados relacionados

Resultados totales (Incluyendo duplicados): 8
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

WeGREENnovate, nuevo proyecto financiado en el ámbito de la Malherbología, WeGREENnovate: nuevo proyecto para la gestion de malas hiebas mediante innovaciones tecnológicas

  • Peña Barragán, José Manuel
  • Castro, Ana Isabel de
El proyecto WeGREENNOVATE tiene como objetivo principal desarrollar y valorizar estrategias de manejo del cultivo precisas, eficientes y sostenibles mediante la integración de innovaciones tecnológicas y enfoques agroecológicos. Con el fin de optimizar el control de malas hierbas en cultivos específicos y promover la biodiversidad floral mediante el manejo del hábitat, WeGREENNOVATE se alinea con cuatro objetivos clave del Pacto Verde: 1) reducir el uso de herbicidas; 2) fomentar la agricultura ecológica; 3) preservar la biodiversidad; y 4) incrementar las poblaciones de polinizadores. Para lograrlo, WeGREENNOVATE propone el uso de tecnologías, como la teledetección, detección próxima, aprendizaje automático y actuadores de precisión, considerándolas herramientas esenciales para abordar los desafíos propuestos en dicho proyecto., Trabajos financiados por los proyectos PID2023-150108OB-C31, PID2023-150108OB-C32 y PID2023-150108OB-C33 del MCIU/AEI/10.13039/501100011033/ FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Intelligent Inter- and Intra-Row Early Weed Detection in Commercial Maize Crops

  • Gómez, Adrià
  • Moreno, Hugo
  • Andújar, Dionisio
Weed competition in inter- and intra-row zones presents a substantial challenge to crop productivity, with intra-row weeds posing a particularly severe threat. Their proximity to crops and higher occlusion rates increase their negative impact on yields. This study examines the efficacy of advanced deep learning architectures—namely, Faster R-CNN, RT-DETR, and YOLOv11—in the accurate identification of weeds and crops within commercial maize fields. A comprehensive dataset was compiled under varied field conditions, focusing on three major weed species: <i>Cyperus rotundus</i> L., <i>Echinochloa crus-galli</i> L., and <i>Solanum nigrum</i> L. YOLOv11 demonstrated superior performance among the evaluated models, achieving a mean average precision (mAP) of 97.5% while operating in real-time at 34 frames per second (FPS). Faster R-CNN and RT-DETR models achieved a mAP of 91.9% and 97.2%, respectively, with processing capabilities of 11 and 27 FPS. Subsequent hardware evaluations identified YOLOv11m as the most viable solution for field deployment, demonstrating high precision with a mAP of 94.4% and lower energy consumption. The findings emphasize the feasibility of employing these advanced models for efficient inter- and intra-row weed management, particularly for early-stage weed detection with minimal crop interference. This study underscores the potential of integrating State-of-the-Art deep learning technologies into agricultural machinery to enhance weed control, reduce operational costs, and promote sustainable farming practices., This research was funded by the grant PID2023-1501080B-C33 from MCIU/AEI/10.13039/501100033 and FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Spatio-temporal stability of intelligent modeling for weed detection in tomato fields

  • Gómez, Adrià
  • Moreno, Hugo
  • Valero, Constantino
  • Andújar, Dionisio
[Context] Site-specific Weed Management (SSWM) represents a shift towards precision and sustainability in agricultural weed control by applying treatments selectively. Leveraging machine learning (ML) and deep learning (DL), particularly convolutional neural networks (CNNs), enhances weed detection capabilities through automated image analysis. Challenges such as requiring extensive labeled datasets and spatio-temporal variability of weeds remain. Utilizing multi-year datasets provides an effective solution by reducing labor-intensive annotation efforts and improving model generalization across varying conditions., [Objective] This study aimed to develop and assess DL models (YOLOv8, YOLOv9, YOLOv10, RT-DETR) for accurate detection of five weed species (Cyperus rotundus L., Echinochloa crus-galli L., Setaria verticillata L., Portulaca oleracea L., Solanum nigrum L.) in tomato fields by integrating multi-year datasets through temporal, spatial, and spatio-temporal analyses., [Methods] The study was conducted in commercial tomato fields in Santa Amalia, Badajoz, Spain, across the 2021 and 2022 growing seasons. Images were systematically captured using a high-resolution Canon camera following an “M”-shaped trajectory. Experts annotated weed species manually, and images underwent augmentation and segmentation to enhance dataset robustness. The models were trained on distinct temporal (year-based), spatial (field-based), and spatio-temporal (combined) datasets, utilizing pre-trained weights and consistent hyperparameters. Performance was evaluated using Average Precision (AP) and Mean Average Precision (mAP)., [Results and conclusions] Applying DL models effectively detected weed species, highlighting the importance of comprehensive datasets for model performance. The temporal analysis, relying solely on previous-year data, achieved a mean average precision (mAP) of 0.7, demonstrating limited efficacy. Using same-year but different-field data, spatial analysis showed improved accuracy (mAP: 0.81–0.89). The spatio-temporal analysis combining datasets from multiple years provided the highest accuracy (mAP up to 0.91), validating the advantages of integrating diverse data across different periods., [Significance] Integrating multi-year datasets substantially enhances the precision and efficiency of DL-based weed detection models. This methodology minimizes the dependency on labor-intensive annotation processes, accelerates model deployment, and boosts dataset variability, thus fostering more robust and accurate predictive capabilities. This analysis ensures a comprehensive representation of environmental conditions, soil types, and weed emergence patterns, extending the method's applicability beyond the original sampled areas in tomato commercial fields. Moreover, the environmental and agricultural conditions of the sampled area resemble those found in other major tomato-producing areas worldwide, thus supporting the reproducibility of the experiments and the broader applicability of the study's findings., This research was funded by the grant PID2023-150108OB-C33 funded by the MICIU / AEI / 10.13039/501100011033 and FEDER, EU, and by the grant TED2021-130031B-I00 funded by MICIU/AEI/ 10.13039/501100011033, as appropriate, by the “European Union NextGenerationEU/PRTR., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Datos de investigación Set de datos (Dataset)

GROUNDBASED_WEED: Ground-Based Imagery for Early Weed Classification in Tomato Crops

  • Moreno, Hugo
  • Rivera Cárdenas, Gabriel Eduardo
  • Andújar, Dionisio
[Description of methods used for collection/generation of data] A handheld Canon PowerShot SX540 HS camera was utilized, offering a spatial resolution of 5184 × 3888 pixels. To ensure optimal image quality under dynamic and variable lighting conditions, the camera operated with an automated ISO sensitivity setting and a shutter speed of 1/1000 second. A zenithal (overhead) image was captured from an elevation ranging between 1 and 1.5 meters., [Methods for processing the data] The annotation of the dataset was performed using LabelImg (version 1.8.6), a free and open-source graphical image annotation tool developed in a Python environment, with the PascalVOC format employed for exporting the corresponding XML files. This format records the coordinates of bounding boxes that delineate the objects within each image. All bounding boxes associated with a single image were stored in one XML file. Each object entry includes the coordinates of the upper-left and lower-right corners—denoted as xmin, ymin, xmax, and ymax—along with the class name of the labeled object. Six plant species were categorized with the assistance of weed science experts and labeled according to the EPPO (European and Mediterranean Plant Protection Organization) code system. Another class was included as NR as unrecognized species due to insufficient image resolution or small plant size. The assigned labels are as follows: 1. SOLNI: Solanum nigrum L. 2. POROL: Portulaca oleracea L. 3. SETVE: Setaria verticillata L. 4. ECHCG: Echinochloa crus-galli L. 5. CYPRO: Cyperus rotundus L. 6. LYPES: Solanum lycopersicum L. 7. NR: Unrecognized species due to insufficient image resolution or small plant size These annotations provide a structured foundation for training and validating object detection models, facilitating accurate weed identification in early growth stages. As aforementioned in DATA & FILE OVERVIEW, the dataset includes annotations in multiple standard formats, namely Pascal VOC XML, COCO JSON, and YOLO, ensuring compatibility with a wide range of computer vision frameworks., Accurate identification of weed species at early developmental stages is essential for advancing precision agriculture. Species-level classification enables the implementation of site-specific management strategies, reducing herbicide usage and fostering sustainable crop production. This study introduces a curated dataset of RGB images captured using a handheld Canon PowerShot SX540 HS camera, which provides a spatial resolution of 5184 × 3888 pixels. The images were collected from commercial tomato fields in Santa Amalia, Badajoz —an important agricultural tomato hub within the Vegas Altas region of Spain., The dataset consists of images collected from various agricultural fields, with a focus on summer commercial tomato crops and their associated weed species. It includes a total of 1,216 images and 20,991 labeled instances, gathered during the growing seasons of 2021 and 2022. Data collection was conducted in May and June of each year. To reflect the temporal separation of the data, the dataset is divided into two folders: TOMATO_1 and TOMATO_2. The TOMATO_1 folder contains 938 images and 9,060 labeled instances corresponding to the 2021 season, while TOMATO_2 comprises 278 images and 11,931 labeled instances from the 2022 season. The dataset includes five agronomically significant weed species, which are classified into two botanical groups: the monocotyledonous species Cyperus rotundus L., Echinochloa crus-galli L., and Setaria verticillata L.; and the dicotyledonous species Portulaca oleracea L. and Solanum nigrum L. In addition to these weed species, the dataset also contains images of the tomato crop Solanum lycopersicum L., All images are stored in JPG format and are individually annotated to indicate the corresponding weed species. Each image is accompanied by a corresponding XML file in PascalVOC format, which contains the annotation data, including bounding box coordinates and class labels for the identified species according to the EPPO (European and Mediterranean Plant Protection Organization) code. This dataset is particularly well-suited for training and evaluating advanced machine learning and deep learning models for early-stage weed detection and classification., State Research Agency (AEI). This research was funded by the grant PID2023-150108OB-C33 funded by the MCIU / AEI / 10.13039/501100011033 and FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Ground-based imagery dataset for early weed classification in tomato crops

  • Moreno, Hugo
  • Rivera Cárdenas, Gabriel Eduardo
  • Andújar, Dionisio
Accurate identification of weed species at early developmental stages is essential for advancing precision agriculture. Species-level classification enables site-specific management strategies, reducing herbicide use and promoting sustainable crop production. This study introduces a curated dataset of RGB images captured using a handheld Canon PowerShot SX540 HS camera, offering a spatial resolution of 5184 × 3888 pixels. The images were collected during May and June of the 2021 and 2022 growing seasons from commercial tomato fields in Santa Amalia, Badajoz, an important agricultural hub in Spain's Vegas Altas region, known for its intensive tomato cultivation and processing facilities. The dataset contains 1217 JPG images and 21,208 labelled instances. It is divided into two subsets: the first includes 938 images and 9060 instances from 2021, while the second comprises 278 images and 11,931 instances from 2022. Each image is manually annotated to identify individual plant species. This dataset is intended for training and evaluating advanced deep learning models, including convolutional neural networks and vision transformers, to enable early-stage weed detection and classification. By making it publicly accessible, the study supports the development of image-based monitoring systems that improve the efficiency, accuracy, and environmental sustainability of precision agriculture practices., This research was funded by the grant PID2023–150108OB-C33 funded by the MCIU / AEI / 10.13039/501100011033 and FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Datos de investigación Set de datos (Dataset)

Deepfake-Weed

  • Moreno, Hugo
  • Gómez, Adrià
  • Rivera Cárdenas, Gabriel Eduardo
  • Andújar, Dionisio
[Description of methods used for collection/generation of data] The artificial weed imagery contained in the Deepfake-Weed (_1 and _2) was generated using Stable Diffusion, a modern generative AI model that was conditioned with a set of real field images from Moreno et al., 2025. These real images depicted weed species naturally occurring in tomato crops, specifically Solanum nigrum L., Portulaca oleracea L., and Setaria verticillata L. They served as the visual foundation from which Stable Diffusion learned the characteristic shapes, colors, and textures of each weed species. Once this reference material was provided, Stable Diffusion was used to synthesize thousands of new, highly realistic images, resulting in an artificial dataset. Each subset was generated using the methodologies described in Moreno et al. (2023) and Gómez et al. (2026) for Deepfake‑Weed_1 and Deepfake‑Weed_2, respectively.Therefore, the Deepfake‑Weed dataset comprises 10,800 artificially generated RGB images designed to support research and industrial developments on DL–based weed detection. The dataset, which is organized into two subsets (Deepfake‑Weed_1 and Deepfake‑Weed_2), contains a total of 12,592 annotated weed instances, considering the European and Mediterranean Plant Protection Organization (EPPO). References Moreno, Hugo; Gómez, Adrià; Altares-López, Sergio; Ribeiro, Angela; Andújar, Dionisio. Analysis of Stable Diffusion-derived fake weeds performance for training Convolutional Neural Networks. https://doi.org/10.1016/j.compag.2023.108324 Hugo Moreno, Gabriel Rivera, Dionisio Andújar. Ground-based imagery dataset for early weed classification in tomato crops. Data in Brief. Volume 63,2025,112249,ISSN 2352-3409,https://doi.org/10.1016/j.dib.2025.112249. Gómez, A., H. Moreno, and D. Andújar. 2026. “ End-to-End Convolutional Neural Networks Training on Stable Diffusion Deepfake Imagery for Precise Early Weed Detection.” Weed Research 66, no. 1: e70060. https://doi.org/10.1111/wre.70060, [Methods for processing the data] The artificial images produced by Stable Diffusion are automatically labeled so they can be used to train object‑detection models. The process begins by converting each artificial image into an ExG‑R representation, a vegetation‑enhancement transformation that highlights green plant tissue and suppresses soil and background information. These ExG‑R images are then thresholded to create binary masks in which vegetation appears in white and the background in black. From these binary masks, an automated labeling system detects individual connected vegetation regions and places bounding boxes around them, producing initial annotations for each weed instance in the artificial image. Due to the fact that automated labeling is not perfect, especially when plants overlap or when leaf shapes produce irregular contours, a final manual correction step was applied. In this step, a human annotator reviewed the automatically generated bounding boxes and refined them as needed to ensure accurate placement and sizing. This semi‑automated workflow significantly accelerated annotation while maintaining high labeling quality. The annotation is under the PascalVOC format employed for exporting the corresponding XML files. This format records the coordinates of bounding boxes that delineate the objects within each image. All bounding boxes associated with a single image were stored in one XML file. Each object entry includes the coordinates of the upper-left and lower-right corners—denoted as xmin, ymin, xmax, and ymax—along with the class name of the labeled object. Three weed species were categorized with the assistance of weed science experts and labeled according to the EPPO (European and Mediterranean Plant Protection Organization) code system. The assigned labels are as follows: 1. SOLNI: Solanum nigrum L. 2. POROL: Portulaca oleracea L. 3. SETVE: Setaria verticillata L. These annotations provide a structured foundation for training and validating object detection models, facilitating accurate weed identification in early growth stages., State Research Agency (AEI), This research was funded by the grant PID2023-150108OB-C33 funded by the MCIU / AEI / 10.13039/501100011033 and FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

End-to-End Convolutional Neural Networks Training on Stable Diffusion Deepfake Imagery for Precise Early Weed Detection

  • Gómez, Adrià
  • Moreno, Hugo
  • Andújar, Dionisio
The integration of optical sensor systems with advancements in artificial intelligence for image processing presents a promising avenue for the broader implementation of site-specific weed management (SSWM). Convolutional neural networks (CNNs) have significantly advanced the field of weed detection. However, their effectiveness remains contingent upon the availability of extensive, manually annotated image datasets, which require considerable time and effort. This study introduces an innovative methodology to overcome these limitations by training detection models exclusively with artificial images generated using stable diffusion (SD), a state-of-the-art generative artificial intelligence model. The research was conducted in commercial tomato fields naturally infested with three weed species: Solanum nigrum L., Portulaca oleracea L. and Setaria verticillata L. An artificial dataset comprising 9000 artificially generated images was developed to train four CNN-based object detection models: YOLOv8, YOLOv9, YOLO-NAS and RetinaNet. The performance of these models was subsequently validated against real-world weed images. The results demonstrated high detection accuracy, with YOLOv8 achieving a mean Average Precision (mAP) of 94.3%, YOLO-NAS reaching 95.5%, YOLOv9 attaining 95.2% and RetinaNet obtaining a mAP of 86.3%. The high quality and variability of the artificial images produced via SD demonstrate the potential of generative models to accelerate traditional data acquisition. This approach provides a scalable and efficient pathway for rapidly deploying weed detection models and adapting them to new crop-weed scenarios. Notably, it offers a means to significantly reduce the effort and resources required for collecting and annotating large real-world datasets, without compromising detection precision., This research was funded by the grant PID2023-150108OB-C33 funded by the MICIU/AEI/10.13039/501100011033 and FEDER, EU., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Fast Forward the Future: What Are the Key Drivers in Intelligent Sensing for Agriculture?

  • Moreno, Hugo
This article belongs to the Special Issue The Future of Artificial Intelligence and Sensor Systems in Agriculture., Driven by the fusion of advanced robotics and Artificial Intelligence (AI), the industry is evolving into a new era of “Agriculture 4.0”. Within the domain of AI, this shift leverages machine learning (ML) and deep learning (DL) to provide artificial perception and decision-making for UAVs, ground-based autonomous vehicles, and robotic systems, effectively redefining traditional farming through “Smart” technology. These smart farming technologies are expected to boost yields, reduce input use, and lower labor demands [1]. This Editorial aims to position the contributions of the articles featured in the Special Issue “The Future of Artificial Intelligence and Sensor Systems in Agriculture” within a broader context. The advancements highlighted in this Editorial illustrate the ongoing transition from manual, experience-based assessment to data-driven diagnostic systems. This shift is paving the way for a new generation of smart farming technologies in which ML and DL models, along with AI-powered sensor systems, play a central and transformative role. Therefore, a diverse set of innovations is presented to introduce new techniques to enable precise monitoring, automated decision-making, and robotic intervention, fundamentally transforming crop management, including weeds and pest management, disease pathology, and yield estimation., The author acknowledges the support and funding by the grant PID2023–150108OB-C33 funded by the MCIU/AEI/10.13039/501100011033 and FEDER, EU., Peer reviewed