Annotated Insect Image Dataset for Training Oriented Bounding Box and Segmentation Models in Morphometric Analysis

This dataset accompanies the manuscript titled "Deep Learning-Based Methods for Automated Estimation of Insect Length, Volume, and Biomass". It contains the annotated image data used to develop, train, and evaluate two complementary deep learning methods for automated insect morphometrics.

The dataset is divided into two primary components corresponding to the methods developed:

1. Oriented Bounding Box (OBB) Dataset:

Purpose: Used for training and evaluating an OBB model for rapid insect length estimation.

Content: Contains 815 images of insects from diverse families within the orders Diptera, Hymenoptera, and Coleoptera.

Image Acquisition: Images were captured using a low-cost, high-resolution DIY microscope, the Entomoscope (Wührl et al., 2024). Multiple focal planes were stacked using Helicon Focus (Helicon Soft Ltd, 2025) to achieve optimal clarity. Specimens were preserved in ethanol.

Format: YOLO OBB format label files.

Metadata: Specimen details and image lists are provided in the supporting information file.

2. Segmentation Dataset:

Purpose: Used for training and evaluating a segmentation model for detailed curvilinear length and volume estimation, demonstrated with Tachinidae (Diptera).

Content: Comprises a total of 1,320 images of several representative tachinid species.

Image Acquisition: Same as the OBB dataset (Entomoscope, Helicon Focus, ethanol preservation).

Annotation Strategy & Structure:  A two-stage annotation strategy was employed, and the dataset is structured to reflect this:

initial_labeling_strategy_820_images/: Data corresponding to the initial annotation approach focusing only on directly visible body parts

final_refined_dataset_500_images/: This is the dataset used for the final model. The key refinement in this stage was the explicit annotation of inferred outlines for body parts partially obscured by wings or legs, where reliable estimation was possible.

Format: YOLO segmentation format label files.

Metadata: Specimen details and image lists are provided in the Supporting Information file.

General Information:

Diversity: For both datasets, images were captured from various views and orientations to ensure model robustness.

This dataset is provided to ensure the reproducibility of our research and to serve as a resource for the broader community working on automated image-based analysis of insects and other biological specimens.