Aug 2026· Agriculture· Vol 16, pp. 1790· 0 citations· 54 references
TL;DR
DC-FEN, a MobileNetV3-based design that models spatial-token relations and channel interactions in parallel and injects them through gated residual fusion is introduced and shows that adding intermediate transfer constraints does not guarantee a stronger student.
Abstract
Plant disease symptoms combine local texture changes with patterns distributed across a leaf, while practical recognition models must remain compact. We introduce DC-FEN, a MobileNetV3-based design that models spatial-token relations and channel interactions in parallel and injects them through gated residual fusion. We also examine output-distribution, direct-feature, and token-relation transfer under same-backbone and heterogeneous teachers. PlantVillage and Plant Pathology 2021 (FGVC8) are evaluated with duplicate-audited, group-aware 70/15/15 splits, an explicit unresolved-leaf sensitivity check, validation-only selection, five training seeds, class-sensitive metrics, and paired seed-wise descriptive summaries. On PlantVillage, the no-additional-attention student, DC-FEN teacher, and DC-FEN joint student obtain macro F1 scores of 96.46±0.91%, 96.90±0.40%, and 96.55±0.25%. On FGVC8, the corresponding scores are 87.29±0.63%, 87.14±0.52%, and 87.20±0.26%. At the prespecified FGVC8 threshold of 0.5, DCAB changed sample-wise F1 by −0.02±0.55 percentage points relative to the unmodified backbone; validation-selected global and label-specific thresholds changed this contrast to +0.28±0.55 and +0.55±0.29 points, while threshold-free macro mAP remained essentially unchanged. A duplicate-audited PlantDoc pressure test reduced frozen-checkpoint accuracy to 30.34±1.10% and 29.57±1.10%, showing that external generalization remains unestablished. A ResNet50 teacher gives logit-only students 97.42±0.51% macro F1 on PlantVillage and 89.82±0.43% sample-wise F1 on FGVC8. After separately weighting the direct and relation terms, the corresponding joint students obtain 97.37±0.56% and 89.94±0.27%, recovering the degradation seen with unit internal weights while remaining close to logit-only transfer. Thus, the study evaluates the benefits and limits of explicit spatial–channel interaction and shows that adding intermediate transfer constraints does not guarantee a stronger student.
Multi-crop disease recognition becomes difficult when visually similar lesion patterns must be identified under a tight parameter budget. This paper reports a compact conference-scale study for the computer-vision and machine-learning track of MLES 2026. A representative 14-class subset covering tomato, cucumber, grape, and apple was constructed from 13,205 images, including 10,272 training images and 2,933 held-out evaluation images. The proposed network couples a lightweight global branch, implemented by a shallow CNN stem followed by a Transformer encoder, with a MobileNetV3- Small local branch for texture-sensitive feature extraction. A learned gating head projects and adaptively fuses global and local evidence before classification. On a single RTX 3060 GPU, the model achieved 99.35% Top-1 accuracy and 100.00% Top-5 accuracy, with macro precision, recall, F1-score, and specificity of 99.36%, 99.39%, 99.37%, and 99.95%, respectively. The model uses only 2.17M parameters, indicating that accurate and deployable visual recognition is possible with a compact dual-branch design. To address reviewer concerns on robustness and component attribution, the revised manuscript additionally reports five-fold cross-validation statistics, single-branch baselines, augmentation ablations, and a freezing-strategy study.
Yang Zhang, Rongrong Gu, Chengyuan Li et al.· International Conference on...· 0 citations
Crop leaf diseases often differ by small lesion regions and similar color-texture patterns, which can limit the feature selectivity of compact CNN backbones when annotated images are insufficient. To address this problem, this study develops an improved ResNet18 for PlantVillage-based leaf disease recognition. The model combines three changes: an SE channel-recalibration unit for emphasizing lesion-related channels, Leaky ReLU for preserving gradient flow when activations are negative, and ImageNet-based transfer learning for faster adaptation to the target classes. Under the same training and evaluation protocol, the proposed network achieved 97.86% accuracy, with precision, recall, and F1-score of 97.63%, 97.41%, and 97.52%, respectively. These values exceeded the original ResNet18 and several commonly used CNN baselines. Ablation experiments showed stepwise gains from the three modifications, and Grad-CAM visualization indicated more concentrated responses over lesion regions. The results suggest that the modified ResNet18 is an effective compact baseline for crop leaf disease recognition, although validation under complex field conditions remains necessary.
Leaf disease diagnosis needs models that are accurate enough for agronomic use yet small enough for constrained computing settings. This study examines a late-attention MobileNetV2 design in which one Convolutional Block Attention Module (CBAM) is inserted between the last MobileNetV2 convolutional map and global average pooling. The experiments use 54,306 controlled-background PlantVillage images spanning 38 classes. Under a uniform saved-model re-evaluation, MobileNetV2 + CBAM obtained 97.17% accuracy and 97.15% weighted F1-score, whereas MobileNetV2 obtained 96.78% and 96.73%. On the converted models, paired testing gave a 0.64-percentage-point accuracy advantage for the CBAM variant (95% CI: 0.31–0.96; exact McNemar p<0.001). The proposed network has 4.02 million parameters, costs 0.604 GFLOPs (about 0.302 GMACs), and yields a 4.07 MiB dynamic-range-quantized TensorFlow Lite file with 96.70% accuracy. Batch-one inference on an Intel i7-11800H CPU with TensorFlow Lite/XNNPACK and eight threads reached a median of 45.42 ms (P95: 102.54 ms), excluding preprocessing. Grad-CAM inspection illustrates both lesion-centered activation and unresolved shared errors. The evidence therefore supports a compact accuracy–cost compromise for the tested conditions, while field robustness, energy use, repeated training runs, and target-device behavior remain open validation requirements.
Emmanuel Udoh, Mohammed Ayoub Alaoui Mhamdi, M. Allili· Electronics· 1 citation
The fine-scale monitoring of plantation cover disappearance and appearance is challenging because these changes are often expressed as weak within-class variations in high-resolution images. This study proposes MLLAForestCD, a three-class pixel-level semantic change-detection network for Eucalyptus plantations. The model uses an MLLA encoder to model the long-range spatial context with efficient linear attention, while a dual-axis change extractor reorganizes paired bi-temporal features through complementary layouts before contextual interaction. Multi-scale fusion then combines semantic cues with boundary-level details. We further construct the Eucalyptus Change Detection Dataset (ECDD), which contains plantation scenes with weak spectral contrast, fragmented boundaries, and directional canopy textures. Under the retained patch-level training/validation split, MLLAForestCD achieves an F1-score of 96.66% and an mIoU of 93.59%. After separate training and evaluation based on WHU-CD, it achieves an F1-score of 97.29% and an IoU of 90.10%; this result reflects performance under an independent WHU-CD training protocol. Finally, annual change maps from 2020 to 2023 are used to derive the most recently detected plantation-appearance time within the observation window. The resulting product is interpreted as a recent stand-renewal event map and requires independent forestry records before biological stand age can be inferred.
Sugarcane yield is reduced by leaf diseases such as mosaic, red rot, rust, and yellow leaf, whose overlapping symptoms make early field diagnosis difficult. This paper presents UniDetNet-MAF, a lightweight unified detection framework that simultaneously localizes and classifies sugarcane leaf diseases. Three components drive its performance: MAF-Conv (a multi-dimensional attention convolution acting jointly on input-channel, kernel-spatial, kernel-count, and output-channel axes), Hierarchical Fusion Block (HFB) and MSAF-Net (a HFB coupled with a multi-scale attention fusion network providing bidirectional feature aggregation), and SAC-Loss (a shape-aware constrained bounding-box regression loss that improves localisation of irregular lesions). Experiments on the publicly available Sugarcane Leaf Disease Dataset, consisting of 2,521 images across five classes, namely Healthy, Mosaic, Red Rot, Rust, and Yellow Leaf, show that UniDetNet-MAF achieves 94.23% accuracy, 95.12% precision, 93.87% recall, 94.51% F1-score, and 93.68% Matthews Correlation Coefficient (MCC), along with mAP@0.5 = 0.937 and 112 FPS on a single GPU, an improvement of roughly 11% over the LeafNet baseline across all classification metrics. The framework is therefore suitable for timely, field-scale sugarcane disease diagnosis.
A. Yashaswini, M. Natesh· Engineering, Technology &...· 0 citations
Introduction Accurate classification of unsound wheat kernels is important for automated grain quality assessment, but improved recognition performance often comes at the cost of increased model complexity. Methods This study presents ECA-ModNet, a parameter-efficient convolutional network derived from EfficientNetV2-S. The architecture replaces two early-stage Fused-MBConv blocks with Mod-FusedMBConv blocks to introduce input-dependent local contextual modulation and replaces the squeeze-and-excitation modules in later stages with efficient channel attention to model local cross-channel interactions using fewer attention-related parameters. Experiments were conducted on the seven-class G600 wheat subset of the GrainSpace dataset. Results Across three independent runs, ECA-ModNet achieved an accuracy of 90.17 ± 0.21% and a macro-F1 score of 90.23 ± 0.21%, improving upon EfficientNetV2-S by 3.80 and 3.84 percentage points, respectively. The parameter count decreased from 20.19M to 16.49M, while FLOPs increased marginally from 2.90G to 2.95G. ECA-ModNet achieved accuracy statistically comparable to that of ConvNeXt-Tiny and InceptionNeXt-T while using substantially fewer parameters, and obtained 3.03–4.55 percentage points higher mean accuracy than six lightweight baselines. Discussion Ablation experiments identified two Stage 1 Mod-FusedMBConv blocks with a 3×3 context kernel as the configuration with the highest mean accuracy among those evaluated. These results indicate that ECA-ModNet offers a favorable accuracy–parameter trade-off for image-based classification of unsound wheat kernels.
Ming Chen, Y. Ning, Xiao-Bo Wen et al.· Frontiers in Plant Science· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
