OccPepSeg-YOLO Dataset v1.0: Instance Segmentation of Occluded Peppers in Field Images

This repository provides the dataset, trained checkpoint, configuration files, custom model definitions, offline augmentation script, and repeated-inference benchmark results supporting the manuscript “OccPepSeg-YOLO for Instance Segmentation of Occluded Peppers in Field Images”. The dataset contains 7,500 RGB field images and corresponding instance-level segmentation annotations for peppers under leaf occlusion, fruit overlap, scale variation, and complex-background conditions. It includes 7,200 training images, 150 validation images, and 150 test images. The package also provides the custom OccPepSeg-YOLO model files, archived training configuration, trained checkpoint, and the raw and summarized RTX 3090 repeated-inference benchmark results.