Yavuz Selim Şahin
Nimet Sema Gençer1†‡
Hasan Şahin- 1Bursa Uludağ University, Faculty of Agriculture, Department of Plant Protection, Bursa, Türkiye
- 2Bursa Technical University, Faculty of Engineering and Natural Sciences, Department of Industrial Engineering, Bursa, Türkiye
Tomato (Solanum lycopersicum L.) cultivation is crucial globally due to its nutritional and economic value. However, the crop faces significant threats from various pests, including Tuta absoluta, Helicoverpa armigera, and Leptinotarsa decemlineata, among others. These pests not only reduce yield but also increase production costs due to the heavy reliance on pesticides. Traditional pest detection methods are labor-intensive and prone to errors, necessitating the exploration of advanced techniques. This study aims to enhance pest detection in tomato cultivation using AI-based detection and language models. Specifically, it integrates YOLOv8 for detection and segmentation tasks and ChatGPT-4 for generating detailed, actionable insights on the detected pests. YOLOv8 was chosen for its superior performance in agricultural pest detection, capable of processing large volumes of data in real-time with high accuracy. The methodology involved training the YOLOv8 model with images of various pests and plant damage. The model achieved a precision of 98.91%, recall of 98.98%, mAP50 of 98.75%, and mAP50-95 of 97.72% for detection tasks. For segmentation tasks, precision was 97.47%, recall 98.81%, mAP50 99.38%, and mAP50-95 95.99%. These metrics demonstrate significant improvements over traditional methods, indicating the model’s effectiveness. The integration of ChatGPT-4 further enhances the system by providing detailed explanations and recommendations based on detected pests. This approach facilitates real-time expert consultation, making pest management accessible to untrained producers, especially in remote areas. The study’s results underscore the potential of combining AI-based detection and language models to revolutionize agricultural practices. Future research should focus on training these models with domain-specific data to improve accuracy and reliability. Additionally, addressing the computational limitations of personal devices will be crucial for broader adoption. This integration promises to democratize information access, promoting a more resilient, informed, and environmentally conscious approach to farming.
Introduction
Tomato (Solanum lycopersicum L.) is a globally significant vegetable crop, essential for both nutritional value and economic stability. However, tomato cultivation faces substantial threats from various pests. Key pests include Tuta absoluta (Lepidoptera: Gelechiidae), which has a significant socioeconomic impact in Eastern Africa due to its widespread distribution and increased costs and pesticide use among farmers (Pereyra and Sánchez, 2006; Shaltiel-Harpaz et al., 2015; Aigbedion-Atalor et al., 2019). Helicoverpa armigera (Lepidoptera: Noctuidae) is another critical pest, highlighting the low adoption of biological control measures and underscoring the need for improved farmer knowledge and extension programs (Balipoor and Ommani, 2014). Leptinotarsa decemlineata (Coleoptera: Chrysomelidae) and Bemisia tabaci (Hemiptera: Aleyrodidae) also pose substantial threats. Myzus persicae (Hemiptera: Aphididae), along with Dolycoris baccarum (Hemiptera: Pentatomidae), Phyllotreta spp. (Coleoptera: Chrysomelidae), and Nezara viridula (Hemiptera: Pentatomidae), further complicate tomato cultivation. Additionally, Tetranychus urticae (Trombidiformes: Tetranychidae) is a significant allergen, particularly among greenhouse workers and asthmatics living near orchards (Jee et al., 2000). Frankliniella occidentalis (Thysanoptera: Thripidae) is another pest impacting tomato crops. Insecticide use patterns among tomato farmers in Ghana reveal a mix of recommended and non-recommended, persistent insecticides, highlighting the need for better regulation and education (Danquah et al., 2010).
Efficient and timely identification of pests is essential for maintaining crop health and optimizing yield. Traditionally, this process has relied heavily on human observation, which is labor-intensive, time-consuming, and susceptible to errors (Danquah et al., 2010). Artificial Intelligence (AI) models, which use algorithms and computational power to simulate human intelligence, offer a promising alternative. There are various types of AI models for data processing: some models process images by converting them into matrices (detection models), while others process text by converting characters or tokens into vectors (language models) (Vaswani et al., 2017). Detection models, such as Mask R-CNN, Faster R-CNN, SSD, and YOLO (You Only Look Once), provide rapid and accurate pest detection, significantly reducing the need for manual labor and enhancing precision. They are capable of processing large volumes of data in real-time, thereby greatly improving agricultural efficiency and sustainability (Liu and Wang, 2020; Swinburne et al., 2022; Jin et al., 2022; Wen et al., 2022; Rajamohanan and Latha, 2023).
YOLO excels due to its real-time processing, high detection accuracy, and versatility in both detection and segmentation tasks. Unlike traditional AI-based pest management systems, this study introduces a novel integration of real-time detection with YOLOv8 and language-based decision support via ChatGPT-4, offering both precision in pest detection and actionable, context-specific recommendations for farmers. This combination allows not only for accurate detection but also for informed decision-making, making the system accessible and practical for real-world agricultural applications. By reducing the reliance on manual expertise and providing timely insights, this system improves both the efficiency and sustainability of pest management practices. It is particularly effective for detecting small, densely packed objects like agricultural pests, making it ideal for real-time applications (Redmon et al., 2016). Its adaptability to various scales and high mean Average Precision (mAP) scores further justify its use in training and detecting agricultural pests, effectively managing multiple pest species with diverse morphologies (Yang et al., 2020; Hashimoto et al., 2020). These features make YOLO an excellent choice for pest detection and segmentation in this study.
However, while detection models like YOLO have the potential to analyze pests more accurately and quickly than humans, they lack the capability to interpret the findings and provide actionable recommendations to farmers (Swinburne et al., 2022; Jin et al., 2022). This gap, which requires knowledge and experience, can be filled by language models. Language models, like ChatGPT, are a type of AI designed to understand and generate human language. They process data by converting characters into vectors, which allows the model to recognize and predict patterns in text (Vaswani et al., 2017). ChatGPT-4, developed by OpenAI, was trained on approximately 1.3 trillion tokens, providing it with a vast knowledge base (Rao et al., 2023). Therefore, while YOLO is used for accurate and real-time pest detection, ChatGPT was chosen as the language model for this study due to its extensive training and ability to generate relevant, insightful responses to interpret the detected tomato pests.
Accurate and real-time identification of agricultural pests necessitates education, knowledge, and experience (Swinburne et al., 2022; Danquah et al., 2010). Once pests are detected, it is essential to have detailed information about them to devise effective management strategies (Balipoor and Ommani, 2014; Aigbedion-Atalor et al., 2019). Accessing this information can be time-consuming and costly. However, language models can provide detailed commentary on detected pests in agricultural applications, thus informing farmers who may lack expertise. By facilitating access to accurate information and analyzing large datasets more quickly than humans, these models can save time and costs while enhancing the quality of education. In this study, detection models and language models are integrated through an API (Application Programming Interface, a set of rules and protocols for building and interacting with software applications, allowing different systems to communicate and share data) to analyze and interpret pest data, providing a valuable guide for future similar research endeavors.
Materials and methods
Definition of the research area and dataset
Turkey is one of the top five tomato-producing countries in the world. About 10% of Turkey’s tomato production occurs in Bursa, where tomatoes were the most produced vegetable in 2020, with 13.2 million tons (Kumbasaroğlu et al., 2021). This study was conducted from March 2022 to September 2023 in Bakirköy village, located in the Karacabey district of Bursa province in the northwest of Turkey, lying between latitudes 40°7’17.53”N and 40°10’40.36”N and longitudes 28°21’14.12”E and 28°26’2.37”E. Field campaigns were conducted from June to July 2023. The site covers an area of 47.16 km², and a total of 96 tomato fields were investigated. The identification of pests observed in the field photographs was carried out according to the morphological diagnostic keys available in the literature (Blackman and Eastop, 2000; Hoebeke and Carter, 2003; Desneux et al., 2010; Ashbrook et al., 2022; Li et al., 2023).
Detection models excel in identifying the presence and location of pests quickly and efficiently (Barbedo, 2016). However, segmentation models are more suitable when detailed morphological features or comprehensive damage maps are necessary (Arockia et al., 2023). The YOLOv8 model integrates both detection and segmentation capabilities. In this study, the yolov8s.pt model was employed for detection tasks, while the yolov8n-seg.pt model was utilized for segmentation tasks. The images used for detection and segmentation in this study encompass various pests and damage types affecting tomato crops. These include Dolycoris baccarum (Hemiptera: Pentatomidae), Phyllotreta spp. (Coleoptera: Chrysomelidae), Nezara viridula (Hemiptera: Pentatomidae), Myzus persicae (Hemiptera: Aphididae), Bemisia tabaci (Hemiptera: Aleyrodidae), Leptinotarsa decemlineata (Coleoptera: Chrysomelidae), Tuta absoluta (Lepidoptera: Gelechiidae), Helicoverpa armigera (Lepidoptera: Noctuidae), Liriomyza bryoniae (Diptera: Agromyzidae) damage, Frankliniella occidentalis (Thysanoptera: Thripidae) damage, and Tetranychus urticae (Trombidiformes: Tetranychidae) damage. These pests and damage types were systematically photographed and used to train the YOLOv8 model for accurate detection and segmentation tasks, aiming to enhance the model’s ability to identify and manage multiple pest species effectively.
From March 2023 to September 2024, high-resolution images of tomato plant diseases and pests were captured using a Canon EOS 700D camera with a resolution of 768 × 1024 pixels. To ensure consistency in image quality, all photographs were taken using cameras set to identical resolution settings. Images were taken at distances of 1 meter and 0.2 meters from the leaves, from various angles (Yong et al., 2020). A comprehensive dataset of over 1,000 images was compiled for each pest, documenting different angles and features. These images were then divided into three subsets: 80% for training, 18% for validation, and 2% for testing, as outlined in Table 1.
Table 1. Distribution of the image dataset for model training.
Data preprocessing techniques and applications
Data preprocessing refers to a series of steps undertaken to prepare raw data for analysis or modeling. It is commonly used in data mining, machine learning, statistics, and data analysis to address data deficiencies, noise, and inconsistencies, thereby enabling more effective analysis (Şahin and Topal, 2016; Atalan et al., 2022; Baştürk and Şahin, 2022). For image processing models, preprocessing the dataset involves three steps: image labeling, resizing, and augmentation. Augmentation provides a large amount of training data to learn features and achieve accurate classification on unseen data, preventing issues like overfitting and poor generalization (Redmon and Farhadi, 2018; Rubanga et al., 2020; Güven and Şahin, 2022).
In this study, Python (version 3.11.8) was used for data preprocessing due to its extensive library ecosystem. To enhance the quality of model training, the original images taken under field conditions were augmented using the OpenCV library. During augmentation, transformations were applied to each image, including rotation, cropping, flipping, adding noise, adjusting lighting, and zooming out (Shorten and Khoshgoftaar, 2019). All images were resized to 600 × 600 pixels as required for model training (He et al., 2017). The analyses were conducted in the Spyder IDE, part of the Anaconda distribution, which offers various libraries for scientific computing and data science (Şahin et al., 2020; Yılmaz et al., 2021). Prior to any augmentation, the dataset was divided into training, validation, and test sets (80%, 18%, and 2%, respectively) to ensure that no data leakage occurred during the augmentation process. Augmentation was only applied to the training set to avoid introducing artificial examples into the validation and test sets, which could result in overly optimistic performance estimates (Shorten & Khoshgoftaar, 2019). Specifically, transformations such as rotation, cropping, flipping, adding noise, adjusting lighting, and zooming out were applied only to the training data after the initial dataset split. Labeling was performed using the LabelMe tool (https://github.com/wkentaro/labelme), with two main approaches: pixel-based segmentation for precise boundary definitions and rectangular bounding for approximate location and size.
Model setup and training
YOLOv8 was selected for this study due to its superior speed and efficiency compared to slower yet more accurate models like Faster R-CNN, making it particularly well-suited for real-time agricultural pest detection, where timely decisions are essential for effective pest management (Tang et al., 2021). The YOLOv8 models were trained using the ultralytics library for model loading and training, and the google.colab library for accessing the dataset via Google Drive. Training parameters included over 100,000 epochs (with patience set to 50 to prevent overfitting), a batch size of 16, and an image size of 640 (Table 2). The model’s hyperparameters, including the number of epochs, batch size, and learning rate, were optimized through an iterative process. Early stopping (patience) was employed to prevent overfitting, while cross-validation was used to fine-tune the learning rate and batch size. The optimal values for these parameters were selected based on the model’s performance on the validation set, ensuring robustness and preventing overfitting. Training was conducted on Google Colaboratory, utilizing an Intel Xeon CPU, 12.68 GB RAM, and a Tesla K80 GPU. Both detection and segmentation models were trained in Python on a custom dataset. Instance segmentation models were chosen to precisely identify damage caused by multiple pest species on tomato plants, which is crucial for accurately identifying specific damages on leaves and fruits (Mirhaji et al., 2021; Zhang et al., 2023a, 2023). The YOLO framework used in this study is illustrated in Figure 1.
Table 2. Key parameters were set in Google Colab for the training of the Ultralytics YOLOv8.
Figure 1. YOLO Framework used in this study. Input: Raw images fed into the model. Backbone: Extracts features from images. Neck: Combines and enhances features. Prediction: Predicts pests’ presence and location. Output: Provides detection and segmentation results.
Model evaluation methodology and testing process
The model’s generalization capability was assessed using a pre-allocated dataset: 80% for training, 18% for validation, and 2% for testing. Key performance metrics, including precision, recall, mAP50, and mAP50-95, were calculated during training on both training and validation datasets. Detection and segmentation performances were evaluated at various IOU thresholds using precision (P), recall (R), and mean average precision (mAP). The mAP50 metric refers to the mean average precision at a 50% IOU threshold, indicating the accuracy of the model in identifying objects with at least 50% overlap with ground truth labels. On the other hand, mAP50-95 averages precision over IOU thresholds from 50% to 95%, offering a more comprehensive view of the model’s accuracy across different overlap scenarios, which is particularly important in agricultural contexts where pests may be occluded or vary in size (Li et al., 2023). These metrics provide insights into the model’s ability to handle various sizes and overlaps in real-world agricultural environments. Loss metrics—box_loss, cls_loss, and dfl_loss—were analyzed to identify areas for improvement. The confusion matrix summarized predictions across classes, highlighting correct and incorrect classifications. This comprehensive analysis provided a clear understanding of the model’s strengths and weaknesses. Metrics P, R, mAP50, and mAP50-95 are defined by Table 3.
Table 3. Formulas of key performance metrics for evaluating YOLO models in object detection.
Prompt creation and OpenAI GPT-4 integration
The study used Python and open-source libraries to integrate detection models with OpenAI’s GPT-4 via an API key. Initially, models identified trained objects, which were then linked to GPT-4. A good prompt should be clear, specific, and provide context to guide the AI’s response. Labels were defined as ‘det_labels_str’ and ‘seg_labels_str’. The prompt used in the study was: prompt_str = f”Could you provide a detailed explanation, in academic English, on the methods for controlling {det_labels_str} or {seg_labels_str} and the potential damage they can inflict on related plants, including preventative measures and integrated pest management strategies?”. Text outputs, limited to 250-400 tokens, were visualized with detection results. A ten-step coding sequence enabled the simultaneous operation of segmentation and detection models (Table 4). The workflow, from image capture to ChatGPT-4 output, is depicted in Figure 2.
Table 4. Integration of Ultralytics YOLO and OpenAI GPT-4 using API key.
Figure 2. ChatGPT-4 integration process.
Results
Training and validation loss graphs
In this study, significant improvements were observed during YOLOv8 model training for pest detection. For training metrics, the box_loss decreased from 1.84 to 0.54, cls_loss from 3.48 to 0.37, and dfl_loss from 1.54 to 0.86. Similarly, validation metrics showed a decline: val/box_loss reduced from 1.38 to 0.53, val/cls_loss from 3.45 to 0.31, and val/dfl_loss from 1.30 to 0.90. Training was halted at 749 epochs to prevent overfitting, demonstrating effective learning and performance. For segmentation, the train/box_loss decreased from 1.97 to 0.57, train/cls_loss from 4.04 to 0.37, and train/dfl_loss from 1.56 to 0.86, while validation metrics also improved, with val/box_loss reducing from 1.99 to 0.74, val/cls_loss from 2.65 to 0.37, and val/dfl_loss from 1.45 to 0.92. Training stopped at 372 epochs to avoid overfitting, indicating robust model performance (Figure 3).
Figure 3. Changes in training and validation loss values over epochs for the YOLOv8 model trained for pest detection and segmentation.
Performance evaluation metrics
During YOLOv8 model training, significant improvements were noted across key metrics: precision increased from 0% to 98.91%, recall from 0% to 98.98%, mAP50 from 0% to 98.75%, and mAP50-95 from 0% to 97.72%. Training was halted at 749 epochs to prevent overfitting, demonstrating enhanced accuracy and reliability in object detection (Figure 4A). For segmentation, precision improved from 0% to 97.47%, recall from 0% to 98.81%, mAP50 from 0% to 99.38%, and mAP50-95 from 0% to 95.99%, with training stopping at 372 epochs to avoid overfitting (Figure 4B).
Figure 4. Improvements in performance metrics during YOLOv8 model training for object detection (A) and object segmentation (B).
Confusion matrix analysis
A confusion matrix, essential for evaluating a model’s performance, pinpoints misclassifications and highlights areas for potential improvement. The test set, comprising images of adult insects, nymphs, and larvae across 11 classes, facilitated the computation of the confusion matrix. The YOLOv8 model exhibited high accuracy in detection tasks. Specifically, D. baccarum adults were correctly classified 890 times with 15 misclassifications, N. viridula adults were accurately identified 695 times with 15 errors, and M. persicae adults were correctly classified 998 times. Additionally, B. tabaci adults achieved 1025 correct identifications, and L. decemlineata adults were correctly identified 666 times with no errors (Figure 5). The model’s performance for segmentation on the test set revealed notable outcomes across 8 classes: 550 correct detections of L. bryoniae damage, 345 accurate detections of T. absoluta damage on fruit, 450 precise detections of T. urticae damage on leaves, and 565 correct identifications of healthy tomato leaves (Figure 6).
Figure 5. Confusion matrix illustrating YOLOv8 model’s performance in pest detection across 11 classes.
Figure 6. Confusion matrix illustrating YOLOv8 model’s performance in pest segmentation across eight classes.
Prompt creation and real-time textual response to visual data
The integration of YOLOv8 and ChatGPT-4 showcases the powerful combination of computer vision and natural language processing, enabling expert feedback on visual data. This integration was tested on five pictures from the test set, which were not included during the training phase. The responses to the crafted prompt, [prompt_str = f”Could you provide a detailed explanation, in academic English, on the methods for controlling {det_labels_str} or {seg_labels_str} and the potential damage they can inflict on related plants, including preventative measures and integrated pest management strategies?”] are presented in Figure 7. The trained detection and segmentation models processed the test images in approximately 0.10 seconds, while the integration with ChatGPT-4 provided textual responses within 3.5 seconds via the API. Despite being limited to 250-400 tokens, the ChatGPT-4 responses, while not always fully comprehensive, demonstrated the potential to offer key information.
Figure 7. Feedback from ChatGPT-4 based on object labels detected by YOLOv8.
Discussion
In the literature, numerous detection models such as Mask R-CNN, SSD, Detectron, and MobileNet are capable of identifying objects in photographs using image processing techniques (He et al., 2017; Liu et al., 2016; Girshick et al., 2014; Howard, 2017). However, among these models, YOLOv8 is preferred in this study due to its superior performance in agricultural pest detection (Redmon et al., 2016; Bochkovskiy et al., 2020). These superior results can be attributed to several factors, including the large and diverse dataset used for training, YOLOv8’s advanced architecture which allows for real-time processing with high accuracy, and the application of optimized hyperparameters specific to agricultural pest detection. For the detection task of the YOLOv8, precision increased to 98.91%, recall to 98.98%, mAP50 to 98.75%, and mAP50-95 to 97.72%. For segmentation tasks, precision increased to 97.47%, recall to 98.81%, mAP50 to 99.38%, and mAP50-95 to 95.99%. These results are consistent with other studies, such as the Pest-YOLO model achieving 69.59% mAP and 77.71% recall, and another study using YOLOv8 for small pest detection in field crops reporting an mAP of 84.7% (Khalid et al., 2023). Additionally, a study on pest detection in strawberries using segmented image datasets achieved a pest detection rate of 91.93% and detection reliability of 83.41% (Choi et al., 2022).
The integration of AI-based detection models with language models like ChatGPT offers significant benefits in pest detection and environmentally friendly pest control (Gu et al., 2021). Traditional methods, such as literature reviews, are resource-intensive, whereas language models provide rapid interpretations within 3.5 seconds, as demonstrated in this research. Researchers emphasize ChatGPT’s potential to train producers and improve information access (Ray, 2023; Siche and Siche, 2023). However, challenges exist regarding output accuracy, which depends on the training data (Gaddikeri et al., 2023). Inaccurate training data can compromise response precision, highlighting the need for training with credible sources.
Open-source language models like LLAMA (Meta) (Touvron et al., 2023), GPT-Neo and GPT-J (EleutherAI) (Black et al., 2022), BERT (Hugging Face) (Devlin et al., 2019), and GPT-2 (OpenAI) (Radford et al., 2019) allow for training on local computers with specific, reliable, and targeted datasets. Nevertheless, even with these advanced models, their effectiveness is contingent upon the quality of the input data and their ability to generalize across diverse agricultural environments. Furthermore, the computational power of personal computers may be insufficient for effectively using these models (Brown et al., 2020). While this study employed the YOLOv8 model integrated with the broadly-informed GPT-4 via an API, utilizing models trained with domain-specific, reliable data could enhance the accuracy and reliability of outputs. Future work should focus on training with domain-specific, trustworthy sources to improve accuracy and applicability across various sectors.
Conclusion
The integration of AI-based detection and language models in this study demonstrates a significant advancement in agricultural practices. By embedding these models into common devices like smartphones, even untrained producers can access real-time expert consultation, enabling immediate pest detection and sustainable pest control. This technology holds the potential to revolutionize agriculture, particularly in remote areas, by reducing costs and facilitating integration with unmanned vehicles for continuous monitoring.
The study’s results, showing substantial improvements in detection and segmentation precision, recall, and mAP metrics, underscore the efficacy of YOLOv8 in agricultural applications. Additionally, integrating language models like ChatGPT enhances the system’s capability by providing detailed explanations and recommendations based on detected pests. This combination allows for rapid, informed decision-making, improving pest management strategies.
Future work should focus on training these models with domain-specific, reliable data to further enhance their accuracy and applicability. Moreover, addressing the computational limitations of personal devices for running advanced models will be crucial for broader adoption. To fully realize the potential of this technology in low-income and remote agricultural settings, future work should focus on the development of energy-efficient models that can run on low-power devices and operate under limited connectivity conditions. Additionally, partnerships with local agricultural cooperatives could facilitate the dissemination and training required for widespread adoption. Ultimately, this integration promises to democratize information access, promoting a more resilient, informed, and environmentally conscious approach to farming.
Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.
Author contributions
YŞ: Writing – review & editing, Writing – original draft. NG: Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. HŞ: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.
Funding
The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.
Acknowledgments
We extend our deepest gratitude to Necmettin Kuruhan for graciously allowing us to use his land for the study, and to undergraduate students Doğukan Akkanat and Emre Çakır for their invaluable assistance in data collection.
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
Aigbedion-Atalor, P. O., Hill, M. P., Zalucki, M. P. (2019). The socioeconomic impact of Tuta absoluta (Lepidoptera: Gelechiidae) in Eastern Africa. J. Economic Entomology 112, 1111–1122. doi: 10.1093/jee/toz220
Arockia, V., Epsy, M., Radha, P. (2023). Pest detection using image denoising and cascaded Unet segmentation for pest images. Tuijin Jishu/J. Propulsion Technol. 44 (4), 1359–1371. doi: 10.52783/tjjpt.v44.i4.1040
Ashbrook, A. R., Mikaelyan, A., Schal, C. (2022). Comparative efficacy of a fungal entomopathogen with a broad host range against two human-associated pests. Insects 13, 774. doi: 10.3390/insects13090774
Atalan, A., Şahin, H., Atalan, Y. A. (2022). Integration of machine learning algorithms and discrete-event simulation for the cost of healthcare resources. Healthcare 10, 1920. doi: 10.3390/healthcare10101920
Balipoor, H., Ommani, A. R. (2014). Adoption of biological control measures for Helicoverpa armigera in tomato cultivation. J. Agric. Sci. 6, 56–62.
Barbedo, J. G. A. (2016). A review on the main challenges in automatic plant disease identification based on visible range images. Biosyst. Eng. 144, 52–60. doi: 10.1016/j.biosystemseng.2016.01.017
Baştürk, F., Şahin, H. (2022). Makine öğrenmesi sınıflandırma algoritmaların karşılaştırması: Metinden dil tanıma örneği. Electronic Lett. Sci. Eng. 18, 68–78.
Black, S., Biderman, S., Hallahan, E., Anthony, Q., Gao, L., Golding, L., et al. (2022). Gpt-neox-20b: An open-source autoregressive language model. arXiv [Preprint]. arXiv:2204.06745.
Blackman, R. L., Eastop, V. F. (2000). Aphids on the world’s crops: An identification and information guide. Chichester: Wiley, 466. doi: 10.1016/0022-2011(95)90067-5
Bochkovskiy, A., Wang, C. Y., Liao, H. Y. M. (2020). YOLOv4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.
Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., et al. (2020). Language models are few-shot learners. Adv. Neural Inf. Process. Syst. 33, 1877–1901.
Choi, Y., Kim, N., Paudel, B., Kim, H. (2022). Strawberry pests and diseases detection technique optimized for symptoms using deep learning algorithm. J. Bio-Environment Control 31, 255–260. doi: 10.12791/ksbec.2022.31.3.255
Danquah, O. B., Asare, D. K., Asante, K. P. (2010). Insecticide use patterns among tomato farmers in Ghana. Int. J. Pest Manage. 56, 343–351.
Desneux, N., Wajnberg, E., Wyckhuys, K. A. G., Burgio, G., Arpaia, S., Narváez-Vasquez, C. A., et al. (2010). Biological invasion of European tomato crops by Tuta absoluta: Ecology, geographic expansion and prospects for biological control. J. Pest Sci. 83, 197–215. doi: 10.1016/j.jip.2009.09.011
Devlin, J., Chang, M., Lee, K., Toutanova, K. (2019). “BERT: Pre-training of deep bidirectional transformers for language understanding.” in: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, eds. Burstein, J., Doran, C., Solorio, T. (NAACL). 1, 4171–4186.
Gaddikeri, V., Jatav, M. S., Rajput, J. (2023). Revolutionizing agriculture: Unlocking the potential of ChatGPT in agriculture. Food Sci. Rep. 4, 20–25.
Girshick, R., Donahue, J., Darrell, T., Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 580–587.
Gu, X., Lin, T. Y., Kuo, W., Cui, Y. (2021). Open-vocabulary object detection via vision and language knowledge distillation. arXiv preprint arXiv:2104.13921.
Güven, Ö., Şahin, H. (2022). Predictive maintenance based on machine learning in public transportation vehicles. Mühendislik Bilimleri ve Araştırmaları Dergisi 4, 89–98. doi: 10.46387/bjesr.1093519
Hashimoto, A., Ohtsu, N., Kudo, H. (2020). Adaptability of YOLO model for agricultural pest detection. Comput. Electron. Agric. 168, 105120.
He, K., Gkioxari, G., Dollár, P., Girshick, R. (2017). “Mask R-CNN,” in Proceedings of the IEEE International Conference on Computer Vision. 2961–2969.
Hoebeke, E. R., Carter, M. E. (2003). Halyomorpha halys (Stål) (Heteroptera: Pentatomidae): A new adventive pest of some crops in North America. Proc. Entomological Soc. Washington 105, 225–237.
Howard, A. G. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
Jee, Y. H., Park, Y. C., Kim, J. Y. (2000). The allergenic impact of Tetranychus urticae among greenhouse workers. Ann. Allergy Asthma Immunol. 84, 543–547. doi: 10.1016/S1081-1206(10)62520-3
Jin, X., Gao, J., Zhang, L. (2022). Application of AI in pest detection for sustainable agriculture. Agric. Syst. 193, 103233.
Khalid, S., Oqaibi, H., Aqib, M., Hafeez, Y. (2023). Small pests detection in field crops using deep learning object detection. Sustainability 15, 6815. doi: 10.3390/su15086815
Kumbasaroğlu, H., Özçelik, A., Çelikkol, P. (2021). Growing tomato under protected cultivation conditions: Overall effects on productivity, nutritional yield, and pest incidences. Crops 1, 97–110. doi: 10.3390/crops1020010
Li, D., Li, H. Y., Zhang, J. R., Wu, Y. J., Zhao, S. X., Liu, S. S., et al. (2023). Plant resistance against whitefly and its engineering. Front. Plant Sci. 14. doi: 10.3389/fpls.2023.1232735
Liu, J., Wang, X. (2020). Tomato diseases and pests detection based on improved YOLO V3 convolutional neural network. Front. Plant Sci. 11. doi: 10.3389/fpls.2020.00898
Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., et al. (2016). “SSD: Single shot multibox detector,” in Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I, vol. 14. (Springer International Publishing), p. 21–37.
Mirhaji, H., Soleymani, M., Asakereh, A., Mehdizadeh, S. A. (2021). Fruit detection and load estimation of an orange orchard using the YOLO models through simple approaches in different imaging and illumination conditions. Comput. Electron. Agric. 191, 106533. doi: 10.1016/j.compag.2021.106533
Pereyra, P. C., Sánchez, N. (2006). Effect of two solanaceous plants on developmental and population parameters of the tomato leaf miner, Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae). Neotropical Entomology 35, 671–676. doi: 10.1590/S1519-566X2006000500016
Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., Sutskever, I. (2019). Language models are unsupervised multitask learners. OpenAI Blog 1, 9.
Rajamohanan, R., Latha, B. C. (2023). An optimized YOLO v5 model for tomato leaf disease classification with field dataset. Engineering Technol. Appl. Sci. Res. 13, 12033–12038. doi: 10.48084/etasr.6377
Ray, P. (2023). AI-assisted sustainable farming: harnessing the power of ChatGPT in modern agricultural sciences and technology. ACS Agric. Sci. Technol. 6, 460–462. doi: 10.1021/acsagscitech.3c00145
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. (Las Vegas, NV, USA), 779–788.
Redmon, J., Farhadi, A. (2018). YOLOv3: An incremental improvement. arXiv. doi: 10.48550/arXiv.1804.02767
Rubanga, D., Loyani, L., Richard, M., Shimada, S. (2020). A deep learning approach for determining effects of Tuta absoluta in tomato plants. arXiv.
Şahin, H., Güntürkün, R., Hız, O. (2020). Design and application of PLC controlled robotic arm choosing objects according to their color. Electronic Lett. Sci. Eng. 16, 52–62.
Şahin, H., Topal, B. (2016). The effect of the use of information technologies in businesses on cost and financial performance. Int. J. Eng. Innov. Res. (IJEIR) 5, 394–402.
Shaltiel-Harpaz, L., Gerling, D., Graph, S., Kedoshim, H., Azolay, L., Rozenberg, T., et al. (2015). Control of the tomato leafminer, Tuta absoluta (Lepidoptera: Gelechiidae), in open-field tomatoes by indigenous natural enemies occurring in Israel. J. Economic Entomology 109, 120–131. doi: 10.1093/jee/tov309
Shorten, C., Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of big data 6 (1), 1–48. doi: 10.1186/s40537-019-0197-0
Siche, R., Siche, N. (2023). The language model based on sensitive artificial intelligence - ChatGPT: bibliometric analysis and possible uses in agriculture and livestock. Scientia Agropecuaria 1, 111–116. doi: 10.17268/sci.agropecu.2023.010
Swinburne, T., Yadav, S., Kim, J. (2022). Efficiency of AI models in agricultural pest management. J. Agric. Technol. 32, 245–260.
Tang, Z., Chen, Z., Qi, F., Zhang, L., Chen, S. (2021). “Pest-YOLO: deep image mining and multi-feature fusion for real-time agriculture pest detection,” in 2021 IEEE International Conference on Data Mining (ICDM). (IEEE). 1348–1353. doi: 10.1109/ICDM51629.2021.00169
Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., et al. (2023). Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., et al. (2017). Attention is all you need. Adv. Neural Inf. Process. Syst. 30, 5998–6008.
Wen, C., Chen, H., Ma, Z., Zhang, T., Su, H., Chen, H. (2022). Pest-YOLO: A model for large-scale multi-class dense and tiny pest detection and counting. Front. Plant Sci. 13. doi: 10.3389/fpls.2022.973985
Yang, X., Zhao, B., Lin, F. (2020). High mean average precision scores for agricultural pest detection using YOLO. Comput. Electron. Agric. 175, 105456.
Yılmaz, M., Şahin, H., Yıldız, A. (2021). Sectoral application analysis of studies made with deep learning models. Electronic Lett. Sci. Eng. 17, 126–140.
Yong, A., Sun, C., Jun, T. (2020). Research on recognition model of crop diseases and insect pests based on deep learning in harsh environments. IEEE Access 8, 171686–171693. doi: 10.1109/ACCESS.2020.3025325
Zhang, L., Ding, G., Li, C., Li, D. (2023a). DCF-Yolov8: An improved algorithm for aggregating low-level features to detect agricultural pests and diseases. Agronomy 13, 2012. doi: 10.3390/agronomy13082012
Keywords: pest detection, precision agriculture, ChatGPT, YOLOv8, sustainable agriculture
Citation: Şahin YS, Gençer NS and Şahin H (2025) Integrating AI detection and language models for real-time pest management in Tomato cultivation. Front. Plant Sci. 15:1468676. doi: 10.3389/fpls.2024.1468676
Received: 22 July 2024; Accepted: 23 October 2024;
Published: 21 February 2025.
Edited by:
Yongliang Qiao, University of Adelaide, AustraliaReviewed by:
Yinyan Shi, Nanjing Agricultural University, ChinaJian Zhang, University of British Columbia, Canada
Elio Romano, Centro di ricerca per l’Ingegneria e le Trasformazioni agroalimentari (CREA-IT), Italy
Copyright © 2025 Şahin, Gençer and Şahin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Yavuz Selim Şahin, eWF2dXpzZWxpbXNhaGluQHVsdWRhZy5lZHUudHI=
†These authors have contributed equally to this work
‡ORCID: Yavuz Selim Şahin, orcid.org/0000-0001-6848-1849
Nimet Sema Gençer, orcid.org/0009-0007-2435-2384
Hasan Şahin, orcid.org/0000-0002-8915-000X