Digital McPhail Trap — olive fruit fly counting validation image set

This dataset supports the manuscript "Design and Field Evaluation of a Low-Cost, Solar-Powered IoT Node for Digital McPhail Traps in Olive Groves" (submitted to Electronics, MDPI). It contains the 125 daytime McPhail-trap images used in the blind expert re-count that validates the study's olive fruit fly (Bactrocera oleae) counting pipeline, together with the per-image counts. The images are a blind, stratified random sample drawn across the six traps and across catch-density strata, from the 2022 field season on Lefkada Island, Greece. Contents images/ — 125 images (IMG001.jpg … IMG125.jpg).counts.csv — per-image counts and metadata, with columns:image_file — image filename;trap — trap identifier (DT-05 … DT-10);date — acquisition date (YYYY-MM-DD);pipeline_count — count from the DIRT-assisted, human-verified counting pipeline (a daily increment, i.e. the catch newly recorded that day);expert_count — independent blind re-count by a human expert of the total flies visible in the image (the standing, accumulated trap content), performed with the pipeline value hidden;density_bin — catch-density stratum used for sampling (0, 1–5, 6–15, 16–40, 41+).README.txt — dataset description and column definitions.Notes. pipeline_count (daily increment) and expert_count (standing content) measure different quantities; they agree closely at low catch density and diverge at high density as the trap content accumulates over the season. See the recognition-validation section and the validation table of the manuscript for the full agreement analysis (MAE, RMSE, bias, Pearson r, Spearman ρ, Lin's concordance correlation coefficient and Bland–Altman limits). The images are real field imagery; no synthetic or augmented images are included. License: Creative Commons Attribution 4.0 International (CC BY 4.0).