Giovanna Sannino

Interpretable-by-design architectures, such as Concept-Bottleneck Models (CBMs), are essential for trustworthy AI under regulations like the EU AI Act. While classical CBMs rely on costly expert labels, recent automated methods use vision–language models for concept discovery. However, whether automation preserves interpretability remains an open question. We systematically evaluate four automated methods (LF-CBMs, LaBo, PCBMs, and VLG-CBMs) against an expert-supervised CBM using fine-grained binary classification (Amanita muscaria vs. Boletus edulis) from the FungiTastic dataset. Although automated methods match expert accuracy, they systematically fail with regard to two key interpretability requirements: concept atomicity and instance-level grounding. Specifically, automated concepts conflate multiple morphological properties, yield scores ungrounded in visual evidence, and produce biologically implausible class-concept associations; a specific failure mode, cross-class contamination, is one that the expert-supervised vocabulary avoids by construction, though expert supervision remains subject to its own annotation and validation risks. We characterize four distinct failure modes across these architectures, concluding that automated concept discovery cannot yet substitute domain expert supervision in safety-critical tasks where explanation fidelity is a functional requirement alongside predictive accuracy.

Deep learning has shown promise for cotton leaf disease detection, yet two critical gaps remain. First, most studies rely on a single model (Convolutional Neural Network-CNN or Transformer) and do not explore how to effectively fuse these complementary architectures. Second, eXplainable AI (XAI) methods are often used qualitatively, lacking objective benchmarks to guide method selection. To address these gaps, we evaluate six backbone models, comprising four CNNs (ResNet50, EfficientNet-B0, DenseNet121, and MobileNetV2) and two Vision Transformers (ViT-Base and DeiT-Small), on the Kaggle cotton leaf disease dataset, which contains 1711 images across four classes. We then systematically investigate five CNN–Transformer fusion strategies, namely concatenation, attention, weighted, ensemble, and variance-based fusion, to identify the most effective approach for disease classification. The best-performing individual models are DenseNet121 (92.40% accuracy) and ViT-Base (96.49% accuracy). Classification metrics include accuracy, balanced accuracy, precision/recall, F1-score, Cohen’s kappa, MCC, AUC, bootstrap confidence intervals, and McNemar tests. Computational efficiency (FLOPs, inference time, model size) is also reported. Concatenation fusion achieves the highest performance (accuracy = 99.42%, 95% CI: 98.2–100%, weighted F1 = 0.994, MCC = 0.992). For explainability, we quantitatively compare six XAI techniques, GradCAM, GradCAM++, ScoreCAM, LayerCAM, EigenCAM, and AblationCAM, using the pointing game, IoU, AUC, and localization accuracy. EigenCAM yields the best overall explainability score. This study demonstrates that simple feature concatenation between dual backbones (CNN + Transformer) is highly effective for cotton leaf disease detection and provides a benchmark for XAI method selection in plant pathology.

Deep learning classifiers have achieved high accuracy on plant disease recognition tasks, but their decision-making processes remain opaque. Counterfactual explanations (CFs), minimally modified inputs flipping a classifier’s prediction, reveal which input changes are sufficient to alter the decision. While diffusion-based counterfactual generation has been studied primarily on controlled face datasets (CelebA), its systematic evaluation for fine-grained plant disease classification, particularly under in-the-wild conditions, remains limited. In this work, we apply for the first time DiME (Diffusion Models for CFs) to plant disease classification, evaluating its behavior on both controlled (PlantVillage) and in-the-wild (PlantWild) data. We adapt the pipeline using Stable Diffusion with LoRA fine-tuning as the generative backbone, and we propose Plant Verification Accuracy (PVA), a domain-adapted variant of the Face Verification Accuracy (FVA) metric, to measure species identity preservation in the generated counterfactuals. We compare DiME against three established baselines: Wachter (pixel-space gradient), xGEM+, and DiVE (both based on Variational Autoencoders). On PlantVillage, DiME achieves a 7.1 percentage-point PVA drop versus 60–65 pp for Variational Autoencoders baselines while preserving the target flip rate. Extension to PlantWild yields larger PVA drops (16.6–18.6 pp) and qualitative degradation on in-the-wild imagery, identifying current limitations of the approach when applied outside controlled settings.