Mind the Gap: Bridging Prior Shift in Realistic Few-Shot Crop-Type Classification

2025-11-20

AI Summary

In real-world agricultural scenarios, crop classification suffers from both long-tailed class distributions and severe label scarcity. Manually balancing the training set induces train-test distribution mismatch, critically undermining model generalization. To address this, we propose Dirichlet Prior Augmentation (DPA), the first method to explicitly model unknown target-domain label shift within a few-shot learning framework. DPA characterizes class priors via a Dirichlet distribution and integrates them into feature learning, enabling dynamic, differentiable, prior-driven regularization that adaptively calibrates decision boundaries. Experiments demonstrate that DPA significantly enhances training stability and substantially improves few-shot crop classification accuracy under realistic imbalanced distributions. Our approach establishes a transferable, distribution-robust learning paradigm for long-tailed few-shot vision tasks.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Real-world agricultural distributions often suffer from severe class imbalance, typically following a long-tailed distribution. Labeled datasets for crop-type classification are inherently scarce and remain costly to obtain. When working with such limited data, training sets are frequently constructed to be artificially balanced -- in particular in the case of few-shot learning -- failing to reflect real-world conditions. This mismatch induces a shift between training and test label distributions, degrading real-world generalization. To address this, we propose Dirichlet Prior Augmentation (DirPA), a novel method that simulates an unknown label distribution skew of the target domain proactively during model training. Specifically, we model the real-world distribution as Dirichlet-distributed random variables, effectively performing a prior augmentation during few-shot learning. Our experiments show that DirPA successfully shifts the decision boundary and stabilizes the training process by acting as a dynamic feature regularizer.

Problem

• Addressing class imbalance and distribution shift in crop-type classification
• Overcoming scarce labeled data limitations in agricultural applications
• Bridging training-test distribution mismatch for real-world generalization

Innovation

• Models real-world distribution as Dirichlet random variables
• Simulates unknown label distribution skew during training
• Acts as a dynamic feature regularizer for stability