Supriadi Rustad

Early detection of plant leaf diseases is critical for minimizing crop losses and supporting precision agriculture. While Convolutional Neural Networks (CNNs) have demonstrated high accuracy in image-based diagnosis, conventional architectures may not optimally balance spatial localization and channel-wise feature refinement, particularly in multi-crop classification settings. This study proposes a redundancy-aware dual-attention architecture, termed ATSA-DenseNet, which integrates the spatial branch of the Convolutional Block Attention Module (CBAM-Spatial) with Efficient Channel Attention (ECA) within a DenseNet121 backbone. Unlike prior dual-attention frameworks that retain full CBAM and introduce channel-level redundancy, the proposed design isolates complementary spatial and channel mechanisms to improve representational efficiency without increasing computational complexity. The framework is evaluated on controlled multi-crop PlantVillage-derived datasets comprising tomato, potato, pepper, and maize. Across both 3-crop and 4-crop configuration, ATSA-DenseNet consistently outperforms baseline DenseNet121 and single-attention variants, achieving 99.94% accuracy and 0.9994 macro-F1 on the 4-crop setting while maintaining a lightweight footprint (6.96M parameters, 2.87G FLOPs). Grad-CAM visualizations indicate improved localization of disease-relevant regions compared to the baseline. While results are obtained under controlled imaging conditions, the findings demonstrate that redundancy-aware dual-attention enhances feature discrimination efficiency in multi-class agricultural classification tasks. Future work will extend validation to real-field datasets with natural variability.

Surface morphology influences how multiple adsorbed species interact on metal surfaces. This study investigates the coadsorption of hydrazine (N2H4) and hydroxyl (OH) on vicinal Ni(533) surface using density functional theory (DFT) calculations. We identify stable N2H4–OH coadsorption configurations and characterize them by their respective relative energies and structural properties. The most energetically favorable configuration features cis-N2H4 and OH respectively adsorbed on terrace and step-edge sites. In contrast to vicinal surfaces with narrower terrace, such as Ni(211), the wider terrace on Ni(533) limits the formation of extended hydrogen-bonded arrangements. This results in a different energetic ordering and increases the energy separation between the coadsorption configurations. These findings demonstrate how the terrace width can be utilized to control the conformational preference of N2H4 in N2H4-OH coadsorption system.

Tomato is one of the widely available horticultural products and holds significant economic value in Indonesia. However, its productivity is often disrupted by various leaf diseases. This study aims to compare the performance of three CNN architectures—DenseNet121, Xception, and MobileNetV2—in classifying tomato leaf diseases. The dataset used consists of 10,000 balanced images across ten classes: Bacterial Spot, Septoria Leaf Spot, Early Blight, Late Blight, Mosaic Virus, Yellow Leaf Curl Virus, Leaf Mold, Target Spot, Spider Mites Two-Spotted Spider Mite, and Healthy. All images were resized to 224x224 pixels and divided into 80% training data and 20% test data. Augmentation techniques were applied to balance the data across classes. Experimental results show that the Xception architecture outperforms the other models, achieving an accuracy of 98.79%, with a precision of 98.80%, recall of 98.79%, and an F1-Score of 98.78%. These findings indicate that the Xception model is highly effective for plant disease classification and is suitable for implementation in environments with limited resources.