Corn is shown as one of the main food sources among agricultural products. One of the important issues encountered in corn plants is the breeding process. In breeding studies, determining the haploid and diploid types of seeds is of great importance. Traditionally, expert analysis is used to determine the haploid and diploid corn varieties in food products. However, expert analysis is time-consuming, subjective, and expensive. Computer-based vision solutions are a more reasonable requirement for agricultural applications. In this study, haploid and diploid corn were classified with high accuracy and at minimum cost using active learning approaches. Compared to traditional classification approaches, similar classification performance was achieved with %66 less labeled corn samples. Support vector machines were used to identify the least number of corn samples with the highest information. With the performed study, the time and money costs in the labeling process of corn samples were reduced to a minimum.

Turkey is the major producer of dried figs in the world and lives mold-based problems on dried figs production. Among the various types of molds, Aspergillus niger type molds are the most common one on the figs which is called “black mold” in food industry. In fig processing industry, black mold infected figs are manually detected by workers using visual observation and also by a native method which is called “nailing”. The nailing method is labor intensive and it carries the risk of transmitting the black mold to the sound figs. In this study, a hyperspectral imaging system is proposed to detect the black-mold infected figs by using active learning. In food safety application, obtaining a number of labeled samples from opposite classes is usually required to train the classifier. However, labeling the samples before training the classifier includes the risk of labeling the samples, which do not carry significant information for the classifier. In this study, in contrast to labeling before training, an active learning scheme is used to detect the samples which are relevant for the classifier are detected and labeled. Using the proposed system, a classifier with higher accuracy is trained with respectively less number of samples.

Background/Objectives: Reliable uncertainty quantification (UQ) is a prerequisite for deploying automated systems in safety-critical medical imaging workflows, yet existing approaches either sacrifice computational efficiency or provide poorly calibrated confidence estimates. We present UQ-Mamba, a lightweight architecture that embeds uncertainty quantification natively into a Mamba state space model via linearized error propagation. Methods: UQ-Mamba yields per-prediction approximate epistemic and aleatoric uncertainty estimates in a single deterministic forward pass at only 9.5% additional inference overhead. We note that these components are heuristic approximations derived under three explicit assumptions (diagonal covariance, first-order linearization, and scalar mean-activation reduction) and have not been empirically validated as true Bayesian posteriors. By propagating learnable log-variance parameters through the SSM state transition matrix, UQ-Mamba bridges the gap between parameter efficiency and principled calibration without requiring stochastic sampling or multiple forward passes. Results: Evaluated across four medical imaging modalities—CT organ classification, colorectal histopathology, dermoscopy, and chest radiography—UQ-Mamba achieves 89.71% accuracy with ECE = 0.0217 on OrganMNIST using only 466K parameters ($3.3\times$ lower ECE than ResNet-50 at $50\times$ fewer parameters; note that UQ-Mamba optimizes NLL, whereas ResNet-50 uses standard cross-entropy, which is a confounding factor in the ECE comparison), improves the Mamba baseline by 2.42 percentage points on PathMNIST (ECE = 0.1188 after temperature scaling), achieves 68.88% test accuracy with ECE = 0.0597 on HAM10000 dermoscopy (matching EfficientNet-B0 at $9\times$ fewer parameters), and reaches mAUC = 0.8196 on CheXpert chest radiographs. Conclusions: Ablation studies confirm that the SSM propagation mechanism is necessary for meaningful uncertainty decomposition. These results establish uncertainty-aware SSMs as a promising proof-of-concept direction for calibrated, parameter-efficient medical image classification, with potential relevance to resource-constrained deployment settings pending further clinical validation.

Tarım ürünleri arasında mısır bitkisi başlıca besin kaynakları içerisinde gösterilmektedir. Mısır bitkilerinde karşılaşılan önemli konulardan birisi ıslah sürecidir. Islah çalışmalarında tohumların haploid ve diploid tiplerinin belirlenmesi büyük önem taşımaktadır. Geleneksel olarak, gıda ürünlerinde haploid ve diploid mısır türlerinin saptanması için uzman analizi kullanılır. Uzman analizi, zaman alıcı, sübjektif ve pahalıdır. Bilgisayarlı görü tabanlı çözümler, tarımsal uygulamalar için oldukça makul gereksinimlerdir. Bu çalışmada, haploit ve diploit mısırların aktif öğrenme kullanılarak saptanması için bir görüntüleme sistemi önerilmiştir. Geleneksel yaklaşımların aksine, sınıflandırıcı ile ilgili örneklerin algılanması ve etiketlenmesi için aktif bir öğrenme şeması kullanılır. Önerilen sistem kullanılarak, daha az sayıda örnekle daha yüksek doğruluğa sahip bir destek vektör sınıflandırıcı eğitilmiştir.