cs.LGMar 13, 2026
SaveDirPA: Addressing Prior Shift in Imbalanced Few-shot Crop-type Classification
Organizations: Technical University of Munich (TUM), TUM School of Engineering and Design, Department of Aerospace and Geodesy, Chair of Remote Sensing Technology, Munich, 80333, Germany · Technical University of Munich (TUM), Munich Data Science Institute (MDSI), Garching, 85748, Germany · ELLIS Unit Jena, University of Jena, Jena, Germany, Jena, 07743, Germany
Abstract
Real-world agricultural monitoring is often hampered by severe class imbalance and high label acquisition costs, resulting in significant data scarcity. In few-shot learning (FSL) -- a framework specifically designed for data-scarce settings -- , training sets are often artificially balanced. However, this creates a disconnect from the long-tailed distributions observed in nature, leading to a distribution shift that undermines the model's ability to generalize to real-world agricultural tasks. We previously introduced Dirichlet Prior Augmentation (DirPA; Reuss et al., 2026a) to proactively mitigate the effects of such label distribution skews during model training. In this work, we extend the original study's geographical scope. Specifically, we evaluate this extended approach across multiple countries in the European Union (EU), moving beyond localized experiments to test the method's resilience across diverse agricultural environments. Our results demonstrate the effectiveness of DirPA across different geographical regions. We show that DirPA not only improves system robustness and stabilizes training under extreme long-tailed distributions, regardless of the target region, but also substantially improves individual class-specific performance by proactively simulating priors.
Figures & tables
Figure 1: Visualization of a label (prior) distribution shift between train dataset and a long-tailed test dataset.
| Country | # Crop classes | # Parcels |
| Austria | ||
| Belgium | ||
| Flanders | ||
| Wallonia | ||
| Estonia | ||
| Germany |
Table 1: Dataset details of EuroCropsML 2.0 . The # Crop classes column refers to the number of unique crop types in HCATlevel . The # Parcels column refers to unique parcel geometries after removing potential duplicates from the EuroCrops reference data and filtering out clouds, cirrus, snow, and ice.
Figure 2: Pre-training and fine-tuning country split. Countries colored in blue are used for pre-training, whereas those colored in green are used for fine-tuning. The darker the coloring, the denser the labeled agricultural fields.
Figure 3: EuroCropsML 2.0 binned class distribution for Germany (LSand NRW) and Austria, shown on a logarithmic scale, with the pasture meadow grassland grass class removed.
Figure 4: The number of annotated crop classes that are shared and distinct between the pre-training countries—Austria and Germany—and the respective fine-tuning country.
Figure 5: Biogeographical regions present in EuroCropsML 2.0 , as defined by the ( European Environment Agency (EEA), 2016 ) .
| Country | # Crop classes | |||
| Train (validation test) | Validation | Test | Validation test | |
| Belgium | ||||
| Estonia | ||||
| Latvia | ||||
| Lithuania | ||||
| Portugal | ||||
Table 2: Class distribution of the EuroCropsML 2.0 fine-tuning sets, showing the number of distinct crop classes. The total number of classes in the training set equals the number of available crop types and can be extracted from Table 1 .
| Country | Gini coefficient |
| Belgium | |
| Estonia | |
| Latvia | |
| Lithuania | |
| Portugal | |
| Slovakia |
Table 3: Gini coefficient for the test sets of the EuroCropsML 2.0 fine-tuning countries, a measure of dataset skewness. A value of indicates a perfectly balanced class distribution, while a value approaching means almost all samples belong to a single majority class.
| Country | -shot training set | ||||
| Belgium | |||||
| Estonia | |||||
| Latvia | |||||
| Lithuania | |||||
| Portugal | |||||
Table 4: Bhattacharyya distance for fine-tuning countries in EuroCropsML 2.0 , calculated between the fixed test set and the few-shot training sets. The distance is if the two distributions are identical and increases as they become more distinct.
Figure 6: Dirichlet density for classes, defined over the -simplex. The plots show different concentration parameters , where ( Reuss et al., 2026a ) .
Figure 7: Visualization of four possible draws of the pseudo-prior for classes at consecutive training iterations . The orange sampling ( ) produces sparse draws, with a different class dominating at each iteration. With (teal), the draws remain closer to the uniform distribution, indicated by the dashed line. Resampling the pseudo-prior at each iteration exposes the model to a diverse set of imbalance patterns without assuming any fixed test distribution.
Table 5: Few-shot classification results for Belgium . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black . The best-performing PRESTOmodel is highlighted in bold teal .
Table 6: Few-shot classification results for Estonia . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black . The best-performing PRESTOmodel is highlighted in bold teal .
Table 7: Few-shot classification results for Latvia . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black .
Table 8: Few-shot classification results for Lithuania . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black .
Table 9: Few-shot classification results for Portugal . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black .
Table 10: Few-shot classification results for Slovakia . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black .
Table 11: Few-shot classification results for Slovenia . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black . The best-performing PRESTOmodel is highlighted in bold teal .
Table 12: Few-shot classification results for Spain . We show the results for the most-granular crop types, as presented in Table 2 , as well as for the broader land cover classes (HCATlevel ). We report the metrics’ means standard deviations on the test set for each variant and few-shot task, computed over five runs, cf. Section 5 . The best result for each few-shot scenario is highlighted in bold blue . Depending on which loss function achieves the best result, the top result of the other loss function is shown in bold black . The best-performing PRESTOmodel is highlighted in bold teal .
Figure 8: Performance gains achieved by applying DirPA, visualized relative to the Bhattacharyya distance ( cf. Table 4 ) and Gini coefficient ( cf. Table 3 ). Circular markers denote performance gains, while crosses represent losses of the DirPAmethod. The marker sizes indicate the magnitude of gains or losses compared to the respective baseline. An increasing is equivalent to a decreasing number of in the few-shot scenario and represents a bigger prior shift. For Lithuania and Spain, stays constant since the relative class distribution remains the same in each few-shot task.
Figure 9: Performance gains on HCATlevel achieved by applying DirPA, visualized relative to the Bhattacharyya distance ( cf. Table 4 ) and Gini coefficient ( cf. Table 3 ). Circular markers denote performance gains, while crosses represent losses of the DirPAmethod. The marker sizes indicate the magnitude of gains or losses in metrics compared to the respective baseline. An increasing is equivalent to a decreasing number of in the few-shot scenario and represents a bigger prior shift. For Lithuania and Spain, stays constant since the relative class distribution remains the same in each few-shot task.
Figure 10: Comparison of predictive output space on the -simplex for a three-class classification . The Softmax function reduces feature information to a single coordinate on the simplex, producing a point estimate. In contrast, the Dirichlet distributionmaps features to a probability density over the simplex, thereby capturing epistemic uncertainty.
Appendix figures & tables16 assets
Supplementary material from the paper’s appendix.
Appendix
| Node(s) | GPU | VRAM | CPU | vCPUs | Architecture |
| 1–3 | NVIDIA RTX A6000 | GB | Intel Xeon | Ampere | |
| 4 | NVIDIA A100-SXM4 | GB | AMD EPYC | Ampere |
Table 13: Hardware specifications.
| hyperparameter | value / value range |
| training/tuning epochs | |
| tuning trials per setup | |
| tuning warmup steps (batches) | |
| early stopping patience (epochs) | |
| batch size | |
| learning rate | \left[{10}^{-4}{10}^{-2}\right] |
Table 14: Pre-training hyperparameters. The hyperparameters utilized during pre-training, along with their corresponding valid values.
| hyperparameter | value / value range |
| training/tuning epochs | |
| tuning trials per setup † | |
| tuning warmup steps (batches) | |
| early stopping patience (epochs) | |
| batch size | |
| head learning rate | \left[{10}^{-6}{10}^{-2}\right] |
Table 15: General fine-tuning hyperparameters. The hyperparameters utilized during fine-tuning, along with their corresponding valid values.
| hyperparameter | value / value range |
| DirPAconcentration | |
| DirPAscaling factor |
Table 16: DirPAfine-tuning hyperparameters. The DirPAhyperparameters utilized during fine-tuning, along with their corresponding valid values.
Table 17: Median of selected optimal values of DirPAhyperparameters with cross-entropy loss, calculated based on finding the optimal hyperparameter values to maximize Cohen’s kappa. The median is calculated over five Optuna studies with different random seeds.
Table 18: Median of selected optimal values of DirPAhyperparameters with focal loss, calculated based on finding the optimal hyperparameter values to maximize Cohen’s kappa. The median is calculated over five Optuna studies with different random seeds.
Table 19: Relative importance of DirPAhyperparameters with cross-entropy loss, calculated based on finding the optimal hyperparameter values to maximize Cohen’s kappa. The values represent the median over five Optuna studies with different random seeds. The relative importance is calculated based on a total of five tunable hyperparameters.
Table 20: Relative importance of DirPAhyperparameters with focal loss, calculated based on finding the optimal hyperparameter values to maximize Cohen’s kappa. The values represent the median over five Optuna studies with different random seeds. Please note that an additional hyperparameter (FLfocus parameter ) was taken into account in these studies. Thus, the relative importance of and is calculated against a broader set of parameters (six in total) and, therefore, not directly comparable to the ones in Table 19 .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 21: Fine-tuning runtimes for final training on Belgium . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 22: Fine-tuning runtimes for final training on Estonia . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 23: Fine-tuning runtimes for final training on Lithuania . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 24: Fine-tuning runtimes for final training on Latvia . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 25: Fine-tuning runtimes for final training on Portugal . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 26: Fine-tuning runtimes for final training on Slovakia . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 27: Fine-tuning runtimes for final training on Slovenia . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
| algorithm/loss | benchmark task ( -shot) | ||||
| 1 | 5 | 10 | 20 | 100 | |
| CE | |||||
| CEDirPA | |||||
| FL | |||||
| FLDirPA | |||||
Table 28: Fine-tuning runtimes for final training on Spain . We present the average training time (in minutes) as mean standard deviation over five runs (seeds). The fastest run time per few-shot scenario is marked in bold .
More in this topic
Class-Imbalanced Learning
Class-imbalanced learning develops classifiers that recognize rare classes without sacrificing performance on common ones.
See all 256 papers