One of the most challenging aspects of agricultural pest control is accurate detection of insects in crops. Inadequate control measures for insect pests can seriously impact the production of corn and soybean plantations. In recent years, artificial intelligence (AI) algorithms have been extensively used for detecting insect pests in the field. In this line of research, this paper introduces a method to detect four key insect species that are predominant in Brazilian agriculture. Our model relies on computer vision techniques, including You Only Look Once (YOLO) and Detectron2, and adapts them to lightweight formats—TensorFlow Lite (TFLite) and Open Neural Network Exchange (ONNX)—for resource-constrained devices. Our method leverages two datasets: a comprehensive one and a smaller sample for comparison purposes. With this setup, the authors aimed at using these two datasets to evaluate the performance of the computer vision models and subsequently convert the best-performing models into TFLite and ONNX formats, facilitating their deployment on edge devices. The results are promising. Even in the worst-case scenario, where the ONNX model with the reduced dataset was compared to the YOLOv9-gelan model with the full dataset, the precision reached 87.3%, and the accuracy achieved was 95.0%.

Aquaculture produces more than 122 million tons of fish globally. Among the several economically important species are the Serrasalmidae, which are valued for their nutritional and sensory characteristics. To meet the growing demand, there is a need for automation and accuracy of processes, at a lower cost. Convolutional neural networks (CNNs) are a viable alternative for automation, reducing human intervention, work time, errors, and production costs. Therefore, the objective of this work is to evaluate the efficacy of convolutional neural networks (CNNs) in counting round fish fingerlings (Serrasalmidae) at different densities using 390 color photographs in an illuminated environment. The photographs were submitted to two convolutional neural networks for object detection: one model was adapted from a pre-trained CNN and the other was an online platform based on AutoML. The metrics used for performance evaluation were precision (P), recall (R), accuracy (A), and F1-Score. In conclusion, convolutional neural networks (CNNs) are effective tools for detecting and counting fish. The pre-trained CNN demonstrated outstanding performance in identifying fish fingerlings, achieving accuracy, precision, and recall rates of 99% or higher, regardless of fish density. On the other hand, the AutoML exhibited reduced accuracy and recall rates as the number of fish increased.

Identifying and counting fish are crucial for managing stocking, harvesting, and marketing of farmed fish. Researchers have used convolutional networks for these tasks and explored various approaches to enhance network learning. Batch normalization is one technique that improves network stability and accuracy. This study aimed to evaluate machine learning for identifying and counting pirapitinga Piaractus brachypomus fry with different batch sizes. The researchers used one thousand photographic images of Pirapitinga fingerlings, labeled with bounding boxes. They trained the adapted convolutional network model with batch normalization layers added at the end of each convolution block. They set the training to one hundred and fifty epochs and tested batch sizes of 5, 10, and 20. Furthermore, they measured network performance using precision, recall, and mAP@0.5. Models with smaller batch sizes performed less effectively. The training with a batch size of 20 achieved the best performance, with a precision of 96.74%, recall of 95.48%, mAP@0.5 of 97.08%, and accuracy of 98%. This indicates that larger batch sizes improve accuracy in detecting and counting pirapitinga fry across different fish densities.

The aim of this study was to develop a mathematical model to predict the body weight of tilapia fingerlings using variables obtained through computer vision. A total of 2092 tilapia fingerlings and juveniles, weighing between 10 and 100 g, were fasted for 12 h, anesthetized, weighed, and photographed using an iPhone 12 Pro Max at 33 cm height in a closed container with different bottom colors. Images were segmented using Roboflow’s instance segmentation model, achieving 99.5% mean average precision, 99.9% precision, and 100% recall. From the segmented images, area, perimeter, major axis (MA), minor axis (SA), X and Y centroids, compactness, eccentricity, and the MA/SA ratio were extracted. Seventy percent of the data was used to build the model, and 30% for validation. Stepwise multiple regression (backward selection) was performed, using body weight as the dependent variable. The prediction model was: −17.7677 + 0.0007539(area) – 0.0848303 (MA) – 0.108338(SA) + 0.0034496(CX). The validation model showed similar coefficients and R² = 0.99. The second validation, using observed versus predicted values, also yielded an R² of 0.99 and a mean absolute error of 1.57 g. Correlation and principal component analyses revealed strong positive associations among body weight, area, axes, and predicted values. Computer vision proved effective for predicting tilapia fingerlings’ weight.

As abelhas são polinizadoras essenciais para a conservação da biodiversidade e a manutenção de ecossistemas, especialmente no Cerrado brasileiro. No entanto, a identificação de espécies nativas enfrenta o “impedimento taxonômico”, agravado pela escassez de especialistas e pela complexidade morfológica das espécies. Este estudo propõe o uso de redes neurais convolucionais (CNN) para a classificação automatizada de abelhas nativas do Cerrado, com o objetivo de reduzir essa barreira e auxiliar no monitoramento e conservação desses polinizadores. Foram utilizadas imagens de oito espécies, coletadas a partir de plataformas de ciência cidadã e registros próprios, processadas e rotuladas com caixas delimitadoras. A arquitetura YOLOv8 foi treinada em duas bases de dados: uma inicial desbalanceada e uma segunda balanceada com técnicas de aumento de dados (data augmentation). A segunda base, com 1.527 imagens e aproximadamente 480 anotações por classe, permitiu alcançar uma precisão de 79%, recall de 62% e F1-score de 69%. Em um experimento adicional com duas espécies morfologicamente similares (Trigona hyalinata e Trigona spinipes), o modelo atingiu acurácia superior a 79% e F1-score de 90%. Os resultados demonstram a viabilidade do uso de CNNs para a identificação precisa de abelhas nativas, mesmo com bases de dados limitadas, oferecendo uma ferramenta promissora para apoio à pesquisa e conservação da biodiversidade.

This study introduces AgroInsect, a smartphone-based system that combines on-device deep learning detection using the YOLOv11n model, deployed in the TensorFlow Lite (TFLite) format, with geostatistical mapping to support in-field pest monitoring in soybean and maize crops. The system detects four key pest species (Diabrotica speciosa, Dalbulus maidis, Diceraeus spp., and Spodoptera frugiperda) directly on the device, automatically extracts geolocation from image EXIF metadata, validates spatial consistency against farm boundaries using the Haversine formula, and synchronizes detections to a cloud database. Heatmaps and continuous incidence surfaces are generated via Ordinary Kriging (implemented in PyKrige), enabling visualization of spatial pest patterns. Field evaluations on a mid-range Android smartphone demonstrated strong detection performance, achieving 95.1% overall accuracy, F1-scores > 0.94 for all species, and a false detection rate of only 1.1%. Kriging validation under dense sampling yielded R² > 0.94, faithfully reproducing species-specific ecological distributions (localized for Diabrotica speciosa, diffuse for Dalbulus maidis.). The fully integrated, low-connectivity workflow proved robust in real field conditions, positioning AgroInsect as a scalable, practical tool for precision pest monitoring and faster IPM decision-making in resource-limited rural areas.