IP102

该数据集是一个大规模的基准数据集，包含了102种害虫物种的图片。该数据集面临的挑战包括类内差异大、受损农作物的图片、在嘈杂背景下的小尺度昆虫，以及显著的类别不平衡问题。数据集被划分为训练集、验证集和测试集，规模达到了75,222张图片，涵盖了102个物种。该数据集的任务是对害虫进行分类。

This is a large-scale benchmark dataset containing images of 102 pest species. The dataset faces several challenges, including large intra-class variation, images of damaged crops, small-scale insects in noisy backgrounds, and significant class imbalance. It is divided into training, validation, and test sets, with a total of 75,222 images covering the 102 pest species. The task of this dataset is pest classification.

搜集汇总

背景与挑战

背景概述

IP102是一个大规模昆虫害虫识别基准数据集，包含超过75,000张图像，分为102个类别，具有自然的长尾分布和层次分类法。此外，数据集中还有19,000张图像带有边界框标注，适用于目标检测任务。