2026-10-06T04:22:47Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33289044 2026-10-02T17:04:38Z category_30881 category_35630 category_30599 category_33638 category_35903 portal_1059 item_type_3 month_year_10_2026

Data from: A pilot computer-vision model trained to identify slide-mounted scale insect pests from the family Rhizoecidae Azizah Conerly (24610107) Benjamin Rosen (24610110) Scott Schneider (17362555) Biological sciences Computer vision Agricultural biotechnology not elsewhere classified Biosecurity science and invasive species ecology Machine learning convolutional neural network scale insect Invasive Species Prevention identification tool slide mount Natural History Collections plant quarantine source code Rhizoecidae <p dir="ltr">Using root mealybugs (Hemiptera: Rhizoecidae) as a test case, this study gathered high-resolution extended depth of field images from slide-mounted museum specimens and trained a convolutional neural network (CNN) to identify and distinguish among 16 species that are encountered as agricultural pests in plant quarantine or greenhouses. A pretrained EfficientNetV2-M model was trained using BioEncoder. The dataset contained 1,607 images, with 41 to 239 images per category. Performance was improved through the adoption of augmentation techniques. The purpose of study was to identify best practices for further development of an identification tool to support experts and non-experts alike in the identification of challenging slide-mounted insect taxa.</p><p dir="ltr">Supplementary data includes the following:</p><p dir="ltr">S1 Table. Summary of images used in CNN model testing. Includes the species test folder, unique identifiers for images, model prediction and confidence scores, prediction accuracy, group, and additional notes on specimen condition.</p><p dir="ltr">S2 Table. Total number of Rhizoecidae images from the USNM, CDFA, and FSCA collections captured with the Glissando slide scanner.</p><p dir="ltr">S1 Appendix. Python scripts and a list of all used libraries and their versions. This directory holds all python code and Jupyter Notebooks needed to train, validate, and perform inferencing.</p><p dir="ltr">S2 Appendix. Details on training parameters and augmentations. This directory contains the directories generated by BioEncoder natively or that were added during the development process.</p><p dir="ltr">S3 Appendix. R code for analysis of model confidence by prediction category. Uses S1 Table as the data matrix.</p><p dir="ltr">S4 Appendix. CNN model training results. This directory includes training data, inference, logs, metrics, plots, runs, and weights.</p> 2026-10-02T17:04:38Z Dataset Dataset 10.15482/USDA.ADC/33289044.v1 https://figshare.com/articles/dataset/Data_from_A_pilot_computer-vision_model_trained_to_identify_slide-mounted_scale_insect_pests_from_the_family_Rhizoecidae/33289044 CC0