Image-based Orientation, Sex, and Anatomical Segmentation of Phoridae (Diptera)

This dataset contains 1,281 high-resolution images of Phoridae (scuttle fly) specimens used for the tasks of orientation classification, sex classification, and anatomical body part segmentation, as described in our paper: Automated Specimen Triage for Dark Taxa: Deep Learning Enables Orientation, Sex Identification, and Anatomical Segmentation from Robotic Imaging The images were captured using the DiversityScanner and Entomoscope imaging systems. Each image has been annotated by taxonomic experts for the three tasks. Folder Structure- orientation_classification: Data for 4-class orientation classification (Dorsal, Ventral, Left_Lateral, Right_Lateral).- gender_classification: Data for 3-class gender classification (Male, Female, Undetermined).- body_part_segmentation: Data for 9-class (+ background) semantic segmentation. Data Format Classification TasksInside each classification folder (`orientation_classification`/ and `gender_classification`/), you will find:- images: A folder containing all the images.- labels.xlsx: A file mapping each image to its corresponding label and data split (train, validation, test). Segmentation TaskInside the `body_part_segmentation`/ folder, you will find:- images: A folder containing all the images.- labels: A folder containing single-channel PNG masks. Each mask filename corresponds to an image filename. The pixel values in the masks map to anatomical regions as follows:- 0: background- 1: Head- 2: Thorax- 3: Abdomen- 4: Palps + labella- 5: Antenna- 6: Legs- 7: Wings- 8: Halteres- 9: Genitalia