AgriPath: A Systematic Exploration of Architectural Trade-offs for Crop Disease Classification

TL;DR

This paper systematically compares CNNs, contrastive VLMs, and generative VLMs for crop disease classification, revealing their strengths and weaknesses across different conditions and introducing a new benchmark dataset.

Contribution

It introduces AgriPath-LF16, a comprehensive benchmark dataset, and provides a unified evaluation of three model paradigms under various domain conditions.

Findings

CNNs excel in in-domain accuracy but falter under domain shift.

Contrastive VLMs offer robust, efficient performance across domains.

Generative VLMs show resilience to distributional changes but have failure modes.

Abstract

Reliable crop disease detection requires models that perform consistently across diverse acquisition conditions, yet existing evaluations often focus on single architectural families or lab-generated datasets. This work presents a systematic empirical comparison of three model paradigms for fine-grained crop disease classification: Convolutional Neural Networks (CNNs), contrastive Vision-Language Models (VLMs), and generative VLMs. To enable controlled analysis of domain effects, we introduce AgriPath-LF16, a benchmark of 111k images spanning 16 crops and 41 diseases with explicit separation between laboratory and field imagery, alongside a balanced 30k subset for standardised training and evaluation. We train and evaluate all models under unified protocols across full, lab-only, and field-only training regimes using macro-F1 and Parse Success Rate (PSR) to account for generative reliability (i.e., output parsability measured via PSR). The results reveal distinct performance profiles: CNNs achieve the highest accuracy on in-domain imagery but exhibit pronounced degradation under domain shift; contrastive VLMs provide a robust and parameter-efficient alternative with competitive cross-domain performance; generative VLMs demonstrate the strongest resilience to distributional variation, albeit with additional failure modes stemming from free-text generation. These findings highlight that architectural choice should be guided by deployment context rather than aggregate performance alone.