Tahsin Uygun

Reliable and computationally efficient visual perception is essential for autonomous systems operating in complex livestock barns, where animals and operational objects must be accurately identified under variable illumination and occlusion conditions. This study developed BarnSeg-ECA-Lite, a lightweight instance segmentation model for the simultaneous segmentation of Buffalo Heads, Feed, and Mineral Lick Blocks (MLBs) in real buffalo barn environments. The model was developed based on YOLOv8n-seg by incorporating P2-to-P3 feature assistance, Efficient Channel Attention (ECA), a lightweight neck with GhostConv operations, targeted resampling of challenging instances, and higher-resolution mask supervision. An original dataset of 1703 NoIR images was used for training, validation, and independent testing. BarnSeg-ECA-Lite achieved an mAP@0.50 of 0.9313, mAP@0.50:0.95 of 0.7988, and an F1-score of 0.9300, with 2.7276 million parameters, 10.1108 GFLOPs, 10.3 ms inference time, and 74 FPS. It also achieved mAP@0.50 values of 0.9177, 0.9299, and 0.9464 for Buffalo Head, Feed, and MLB, respectively. These results demonstrate a favorable balance between segmentation accuracy and computational efficiency, providing a practical computer-vision foundation for future livestock and barn-resource monitoring applications, including autonomous feed-pushing robots and intelligent livestock management systems.

Aim of study: To develop a real-time detection system for grape leaf blister mite (Eriophyes vitis (Pagst.)), a significant pest in vineyards, using deep learning models. Early detection of pests and diseases is essential to mitigate agricultural losses, especially considering challenges such as climate change, shrinking agricultural areas, and increasing food demand. Area of study: The study was conducted in vineyard areas located in Tokat, Türkiye, a region known for its extensive viticulture. Material and methods: An Arduino Tiny ML kit and the Edge Impulse platform were utilized to deploy the FOMO (Faster Objects, More Objects) MobileNetV2 0.1 model. The model was trained using a custom dataset of grape leaf blister mite images and validated against expert observations. Real-time detection of blister counts caused by E. vitis was evaluated through comparative metrics. Main results: The trained model achieved a classification success of 90% with an F1 score on the validation dataset and an accuracy of 96.25% on the test dataset. Blister numbers were assessed using both expert observations and the real-time detection system, resulting in RMSE (Root Mean Squared Error) of 1.862 and a MAPE (Mean Absolute Percentage Error) of 15.63%. These results demonstrate the system’s reliability in detecting vineyard pests with high precision. Conclusions: The proposed system provides a practical, cost-effective, and accurate approach for pest detection in vineyards, offering significant potential to enhance pest management strategies and reduce economic losses in agriculture.

The objective of this work was to determine the robustness of a mobile agriculture platform for image capturing in different displacement conditions. Robustness was assessed using the finite element analysis method. The software ANSYS 2022 R2 and SolidWorks 2022 were used to analyze the stress and deformation of the structure in three different scenarios. The results were compared with those of the current literature. The highest stress estimated were 32.774 MPa by ANSYS and 31.550 MPa by SolidWorks. The total estimated deformations were 1.6767 mm by ANSYS and 3.452 mm by SolidWorks. These results indicate that the proposed platform is robust under the studied scenarios.

Global warming caused by climate change causes some problems in agricultural production. One of these problems is the increase in various pest populations. This increase poses a serious threat to agricultural products and significantly negatively affects productivity and quality. Insecticides are commonly used to combat pests. However, most of the time, farmers' lack of knowledge in recognizing pests and understanding their effects results in incorrect and excessive spray applications. While excessive use of insecticides harms human health and environmental pollution, it also increases production costs, causes changes in the genetic structures of pests, causing them to become more resistant, and makes agricultural control difficult. Therefore, early detection of pests and their damage to the plant is extremely important. This study aims to develop an accurate and efficient method to detect damage caused by the tomato leaf miner, Tuta absoluta, on tomato leaves. A dataset comprising healthy and damaged tomato leaves was created. Using a hybrid approach, features were extracted through Convolutional Neural Networks (CNNs) with transfer learning and classified using traditional machine learning techniques. Among the methods evaluated, SVM-Linear achieved the highest accuracy with 97.83%, outperforming other classifiers such as Random Forest with 96.14%, Rotation Forest with 95.89%, and SVM-RBF with 90.70%. These results highlight the potential of combining deep learning-based feature extraction with conventional machine learning for early pest detection. This approach offers a practical solution to reduce the misuse of insecticides and improve pest management strategies, contributing to sustainable agriculture.

Pests significantly negatively affect product yield and quality in agricultural production. Agricultural producers may not accurately identify pests and signs of pest damage. Thus, incorrect or excessive insecticides may be used. Excessive use of insecticides not only causes human health and environmental pollution, but also increases input costs. Therefore, early detection and diagnosis of pests is extremely important. In this study, the effectiveness of the instance segmentation method, a deep learning-based method, was investigated for the early detection of the damage caused by the T. absoluta pest in the leaf part of the tomato plant under greenhouse conditions. An original dataset was created by acquiring 800 healthy and damaged images under greenhouse conditions. The acquired images were labelled as bounding box and automatically converted to a mask label with the Segment Anything Model (SAM) model. The created dataset was trained with YOLOv8(n/s/m/l/x)-Seg models. As a result of the training, the box performance of the proposed YOLOv8l-Seg model was measured as 0.924 in the mAP_0.5 metric. The YOLOv8l-Seg model mask values are, respectively: mAP_0.5, mAP_0.5–0.95, Precision, Recall showed the best performance with values of 0.935, 0.806, 0.956 and 0.859. Then, the YOLOv8l-Seg model, trained with different data input sizes, showed the best performance at 640 × 640 size and the lowest performance with a value of 0.699 in the mAP_0.5 metric in the 80 × 80 size. The same dataset was trained with YOLOv7, YOLOv5l, YOLACT and Mask R-CNN instance segmentation models and performance comparisons were made with the YOLOv8l-Seg model. As a result, it was determined that the model that best detected T. absoluta damage in tomato plants was the YOLOv8l-Seg model. The Mask R-CNN model showed the lowest performance with a metric of 0.806 mAP_0.5. The results obtained from this study revealed that the proposed model and method can be used effectively in detecting the damage caused by the T. absoluta pest.

Bağların kontrolünün, takibinin ve bakımının zamanında doğru bir şekilde yapılması çok önemlidir. Bağ zararlılarıyla mücadelede aşırı pestisit kullanımı, insan sağlığını tehlikeye atmakta ve çevre kirliliğine yol açmaktadır. Ayrıca aşırı pestisit kullanımı ekonomik açıdan düşünüldüğünde işletme giderlerinin artmasına sebep olmaktadır. Bu nedenle bağdaki zararlıların ve hasarlarının zamanında teşhisi çok önemlidir. Zamanında tespiti sağlamaya yardımcı olan yöntemlerden biri derin öğrenmedir. Bu çalışmada, bir derin öğrenme nesne algılama algoritması olan YOLOv8x modeli ile bazı bağ zararlılarının (Salkım güvesi, Trips, Bağ yaprak uyuzu ve İki noktalı kırmızı örümcek) yaprak ve meyve kısmında oluşturduğu hasarın tespitine yönelik çalışma gerçekleştirilmiştir. 7 farklı sınıftan ve 3500 görüntüden meydana gelen veri seti oluşturulmuştur. Oluşturulan veri seti; YOLOv8(n/s/m/l/x) modelleri ile eğitilmiştir. Eğitim sonucunda, YOLOv8x modeli performans değerleri sırayla; mAP0,5, mAP0,5-0,95, Kesinlik (Precision), Duyarlılık (Recall), 0,926, 0,648, 0,892 ve 0,903 şeklinde sonuçlar vermiştir. Aynı veri seti, YOLOv7, DETR ve RTMDet modelleriyle de eğitilerek YOLOv8x modeliyle performans karşılaştırmaları yapılmıştır. Karşılaştırma sonucunda bağlarda belirtilen zararlıların oluşturduğu hasarı en iyi tespit eden YOLOv8x modeli olmuştur.