玉米芯实例分割数据集

1. 简介CornHub 是针对真实田间条件下玉米芯的实例分割数据集，包含 304 张高分辨率 RGB 图像，按分辨率分为三种变体：满员版（2432×2432 像素）、LITE 版（1024×1024 像素）、ULTRALITE 版（512×512 像素）；图像在不同拍摄距离（近景、广角）、光照条件（晴天、阴天）和视点下采集，每个玉米芯均标注为独立灰度遮罩（单通道），标签规则为 0 表示背景，1、2、3…… 依次对应单个玉米芯实例，标注精准区分不同玉米芯个体

1. Introduction
CornHub is an instance segmentation dataset targeting corn cobs under real field conditions, containing 304 high-resolution RGB images divided into three variants based on resolution: Full Version (2432×2432 pixels), LITE Version (1024×1024 pixels), and ULTRALITE Version (512×512 pixels). The images were collected under varying shooting distances (close-up, wide-angle), lighting conditions (sunny, overcast), and viewpoints. Each corn cob is annotated with an independent single-channel grayscale mask, with the labeling rule that 0 represents the background, while 1, 2, 3, ... correspond to individual corn cob instances in sequence to enable precise differentiation between distinct corn cob individuals.

搜集汇总

背景与挑战

背景概述

玉米芯实例分割数据集包含304张田间玉米芯的RGB图像，提供三种分辨率版本（满员版、LITE版、ULTRALITE版），并配有精准的单通道灰度遮罩标注。该数据集适用于农业智能化应用中的实例分割算法研发，如玉米芯识别、计数和生长状态监测。