FruitEnsemble: MLLM-Guided Arbitration for Heterogeneous ensemble in Fine-Grained Fruit Recognition

TL;DR

FruitEnsemble is a novel two-stage framework combining heterogeneous ensemble methods and multimodal large language models to improve fine-grained fruit classification accuracy in challenging scenarios.

Contribution

We introduce a comprehensive fruit dataset and a dynamic inference framework that integrates ensemble learning with MLLM-based expert arbitration for enhanced classification.

Findings

Achieved 70.49% classification accuracy on the new dataset.

Outperformed existing state-of-the-art models in fruit recognition.

Demonstrated robustness in difficult sample classification.

Abstract

Fine-grained fruit classification is a critical yet challenging task in agricultural computer vision, primarily hindered by a severe shortage of high-quality datasets and the high visual similarity between classes. To address these challenges, we first constructed a comprehensive dataset comprising 306 fruit categories with 116,233 samples. Moreover, we propose FruitEnsemble, a practical two-stage dynamic inference framework designed to overcome the generalization limitations of static single-model architectures. In the first stage, FruitEnsemble employs a validation-calibrated weighted ensemble of heterogeneous backbones to generate a robust Top-3 candidate pool. To tackle difficult samples, we introduce an expert arbitration mechanism: when ensemble confidence falls below 0.6, a multimodal large language model (MLLM) is triggered to perform rigorous visual verification by integrating external botanical descriptions using Chain-of-Thought (CoT) reasoning. Furthermore, we optimized the training pipeline with a hard sample-aware joint loss. Extensive experiments demonstrate that FruitEnsemble achieves a classification accuracy of 70.49\% and outperforms existing state-of-the-art models. Our framework provides an efficient, deployment-oriented solution for real-world agricultural visual sorting and quality inspection tasks.