Potato Detection Evaluation Data and Supporting Material

# Potato Detection Evaluation Data and Supporting Materials This repository contains evaluation data and supporting materials for the manuscript **"Robust Approach to Potato Detection in Variable Environments."** The public release focuses on the held-out evaluation data used to assess temporal, regional, camera, and field-condition generalization. ## Contents ```text COMPAG_supporting_data/ |-- README.md |-- data.yaml |-- occ_loss.py |-- main_test_2022/ | |-- images/ | `-- labels/ |-- external_field_2026/ | |-- images/ | `-- labels/ `-- results/ |-- proposed.png |-- yolo11n.png `-- corruption_metrics.csv ``` ### Main test set (2022) - The images were independently collected in June 2022 in Asan, Chungcheongnam-do, South Korea. - The negative images were collected in 2022 in Goryeong, Gyeongsangbuk-do, South Korea, during harvester operation when no potatoes were present. ### External field dataset (2026) - The dataset was collected in late June 2026 using a Logitech C920 webcam mounted on a potato harvester. - It includes changes in acquisition year, camera hardware, and field conditions.