Green Pepper Point Cloud Dataset for Calyx Segmentation

This dataset contains 3D point cloud data of green peppers acquired using an RGB-D camera from a single top-down view under controlled conditions. It includes samples from four commonly consumed varieties: sweet pepper (Oigochu), long green pepper, Cheongyang pepper, and Shishito pepper. Each PCD file is labeled with variety information and preprocessed to remove outliers and normalize coordinates.

For segmentation, each point in the PCD files is labeled as either 0 = pericarp (edible body) or 1 = calyx (non-edible part). The dataset was used to train and evaluate variety-specific 3D deep learning models for segmentation and edible mass estimation.

If you use this dataset, please cite the following paper (currently under review):  A 3D deep learning framework for non-destructive estimation of edible mass in green peppers with attached calyx, Yang et al., 2025.