Locust Abdominal Scan Dataset for Sex Identification: Standardized Laboratory Specimens (Four Genera) and Wild Field Populations with Mixed Developmental Stages

Dataset Description: A Lab-to-Field Benchmark for Automated Locust Sex Identification This dataset is designed to support research in automated sex identification and few-shot domain adaptation for agricultural pest management. We employ a structured "Lab-to-Field" sampling strategy to evaluate model robustness across distinct taxonomic groups, developmental stages, and environmental conditions. The dataset consists of high-resolution (600 DPI) abdominal scan images of locusts (Orthoptera: Acrididae), specifically curated to facilitate feature extraction of the abdominal terminalia. 1. Source Domain: Meta-Training Set (N = 1,475) This domain establishes a baseline for standardized sexual dimorphism under controlled conditions, intended for backbone pre-training or meta-learning tasks. Composition: 855 males and 620 females across four genera: Locusta migratoria, Oxya chinensis, Atractomorpha sinensis, and Diabolocatantops pinguis. Quality Control: Only sexually mature adults with fully differentiated external terminalia were included. Specimens were sourced from laboratory or greenhouse environments to ensure clean, high-fidelity representation learning. 2. Target Domain: Open-Field Evaluation Set (N = 142) Collected from agricultural fields in Taiyuan, Shanxi, China (July–August 2023), this domain simulates real-world deployment challenges through two distinct distribution shifts: Taxonomic Shift (Unseen Genera): Includes Trilophidia-like and Megaulacobothrus-like morphotypes. These genera were entirely absent from the source domain, providing a rigorous test for cross-genus generalization of sexual traits. Domain Shift (Environmental & Developmental Complexity): Includes wild populations of Atractomorpha-like and Diabolocatantops-like morphotypes. This introduces a dual challenge: distinguishing sex amidst mixed developmental stages (nymphs and adults) and varying environmental noise (lighting, specimen degradation). Technical Specifications & Preprocessing Imaging Protocol: Raw scans were captured at 600 DPI using a flatbed scanner. To eliminate systematic illumination bias, specimens were arranged in an interleaved layout (alternating males and females) during scanning. Specimen Preparation: Specimens were preserved at -80°C. Appendages were manually retracted to fully expose the diagnostic abdominal terminalia prior to imaging. Annotations: All images are accompanied by taxonomically verified labels for sex and morphotype alignment. Note: This dataset is part of a research project aiming to enhance the operational efficacy of the Sterile Insect Technique (SIT) and proactive pest monitoring.