A Novel MaxViT Model for Accelerated and Precise Soybean Leaf and Seed Disease Identification
Bibliographic record
Abstract
Timely diagnosis of soybean diseases is essential to protect yields and limit global economic loss, yet current deep learning approaches suffer from small, imbalanced datasets, single-organ focus, and limited interpretability. We propose MaxViT-XSLD (MaxViT XAI-Seed–Leaf-Diagnostic), a Vision Transformer that integrates multiaxis attention with MBConv layers to jointly classify soybean leaf and seed diseases while remaining lightweight and explainable. Two benchmark datasets were upscaled through elastic deformation, Gaussian noise, brightness shifts, rotation, and flipping, enlarging ASDID from 10,722 to 16,000 images (eight classes) and the SD set from 5513 to 10,000 images (five classes). Under identical augmentation and hyperparameters, MaxViT-XSLD delivered 99.82% accuracy on ASDID and 99.46% on SD, surpassing competitive ViT, CNN, and lightweight SOTA variants. High PR-AUC and MCC values, confirmed via 10-fold stratified cross-validation and Wilcoxon tests, demonstrate robust generalization across data splits. Explainable AI (XAI) techniques further enhanced interpretability by highlighting biologically relevant features influencing predictions. Its modular design also enables future model compression for edge deployment in resource-constrained settings. Finally, we deploy the model in SoyScan, a real-time web tool that streams predictions and visual explanations to growers and agronomists. These findings establishes a scalable, interpretable system for precision crop health monitoring and lay the groundwork for edge-oriented, multimodal agricultural diagnostics.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".