RESUMO
Plants encounter a variety of beneficial and harmful insects during their growth cycle. Accurate identification (i.e., detecting insects' presence) and classification (i.e., determining the type or class) of these insect species is critical for implementing prompt and suitable mitigation strategies. Such timely actions carry substantial economic and environmental implications. Deep learning-based approaches have produced models with good insect classification accuracy. Researchers aim to implement identification and classification models in agriculture, facing challenges when input images markedly deviate from the training distribution (e.g., images like vehicles, humans, or a blurred image or insect class that is not yet trained on). Out-of-distribution (OOD) detection algorithms provide an exciting avenue to overcome these challenges as they ensure that a model abstains from making incorrect classification predictions on images that belong to non-insect and/or untrained insect classes. As far as we know, no prior in-depth exploration has been conducted on the role of the OOD detection algorithms in addressing agricultural issues. Here, we generate and evaluate the performance of state-of-the-art OOD algorithms on insect detection classifiers. These algorithms represent a diversity of methods for addressing an OOD problem. Specifically, we focus on extrusive algorithms, i.e., algorithms that wrap around a well-trained classifier without the need for additional co-training. We compared three OOD detection algorithms: (a) maximum softmax probability, which uses the softmax value as a confidence score; (b) Mahalanobis distance (MAH)-based algorithm, which uses a generative classification approach; and (c) energy-based algorithm, which maps the input data to a scalar value, called energy. We performed an extensive series of evaluations of these OOD algorithms across three performance axes: (a) Base model accuracy: How does the accuracy of the classifier impact OOD performance? (b) How does the level of dissimilarity to the domain impact OOD performance? (c) Data imbalance: How sensitive is OOD performance to the imbalance in per-class sample size? Evaluating OOD algorithms across these performance axes provides practical guidelines to ensure the robust performance of well-trained models in the wild, which is a key consideration for agricultural applications. Based on this analysis, we proposed the most effective OOD algorithm as wrapper for the insect classifier with highest accuracy. We presented the results of its OOD detection performance in the paper. Our results indicate that OOD detection algorithms can significantly enhance user trust in insect pest classification by abstaining classification under uncertain conditions.
RESUMO
BACKGROUND: Sampling a small number of participants from an entire country is not straightforward. In this case, researchers reluctantly sample from a single setting or few settings, which limits the generalizability of findings. Therefore, there is a need to design efficient sampling method for small sample size surveys that can produce generalizable results at the country level. METHODS: Data comprised of twenty proxy variables to measure health services demands, structures, and outcomes of 413 districts of Iran. We used two data mining methods (hierarchical clustering method (HCM) and model-based clustering method (MCM)) to create homogenous groups of districts, i.e., strata based on these variables. We compared the internal and stability validity of the methods by statistical indices. An expert group checked the face validity of the methods, particularly regarding the total number of strata and the combination of districts in each stratum. The efficiency of selected method, which is measured by the inverse of variance, was compared with a simple random sampling (SRS) through simulation. The sampling design was tested in a national study in Iran, which aimed to evaluate the quality and costs of medical care for eight selected diseases by only recruiting 300 participants per disease at the country level. RESULTS: MCM and HCM divided the districts into eight and two clusters, respectively. The measures of internal and stability validity showed that clusters created by MCM were more separated, compact, and stable, thus forming our optimum strata. The probability of death from stroke, chronic obstructive pulmonary disease, and in-hospital mortality rate were the most important indicators that distinguished the eight strata. Based on the simulation results, MCM increased the efficiency of the sampling design up to 1.7 times compared to SRS. CONCLUSIONS: The use of data mining improved the efficiency of sampling up to 1.7 times greater than SRS and markedly reduced the number of strata to eight in the entire country. The proposed sampling design also identified key variables that could be used to classify districts in Iran for sampling from these target populations in the future studies.