HIVE: Hierarchical identification in visual entomology for pest recognition

Ecological Informatics 2026 Open access

Read full text
Authors: Minseong Kim, Jinhyeok Choi, Junyoung Park, Olga Frunze, Sungtek Kahng, B.J. Choi, Woochul Kang, Hyung Wook Kwon
DOI: 10.1016/j.ecoinf.2026.103911
Cited by: 0
References: 40

Abstract

Highlights• We present HIVE, a benchmark of 2M pest images annotated across 19 taxonomic levels.• Models trained on HIVE improve real-world pest recognition and capture key traits.• Pretraining on HIVE improves downstream performance across pest benchmarks.• Image embeddings from HIVE-pretrained models enable retrieval of morphologically similar pests.1 HIVE: Hierarchical Identification in Visual2 Entomology for Pest Recognition3 Minseong Kima,1, Jin Hyeok Choib,1, Junyoung Parka,b,c,1, Olga Frunzea, 4 Sungtek Kahnga,e, Byoungjo Choia,d, Woochul Kanga,d, Hyung Wook5 Kwona,b,c,∗

Fields of Study & Subjects

Smart Agriculture and AI Phytoplasmas and Hemiptera pathogens Advanced Neural Network Applications

Publication details

Journal / source
Ecological Informatics
Publisher
Elsevier BV
Volume
97
Pages
103911–103911
Publication date
9 July 2026
Language
EN