Chula-ParasiteEgg-11

我们的Chula-ParasiteEgg-11数据集包含来自粪便涂片样品的11种类型的寄生卵，如图1所示，它们的平均大小大约在15-100微米的区间内。图像是从几种不同的设备获取的，包括带有奥林巴斯BX53显微镜的佳能EOS 70D相机机身，带有尼康Eclipse Ni显微镜的DS-Fi2尼康相机机身，三星Galaxy J7 Prime手机，带有尼康Eclipse Ni 10倍目镜镜头的iPhone 12和13或奥林巴斯BX53设备。这导致不同的分辨率，不同的照明和设置条件。有些图像失焦，有些图像嘈杂，有些图像显示运动模糊已被电动舞台显微镜捕获。各种各样的图像质量和外观会产生各种不同的图像特征。这旨在增强检测模型的鲁棒性。训练和测试数据集将被随机划分。我们希望在数据集发布当天制作1,000和250图像/课程以进行培训和测试。这是同类数据中最大的数据集。标签将采用边界框的形式。

Our Chula-ParasiteEgg-11 dataset contains 11 categories of parasite eggs derived from fecal smear specimens, as shown in Figure 1. Their average size ranges roughly between 15 and 100 micrometers. The images were captured using a variety of devices, including a Canon EOS 70D camera paired with an Olympus BX53 microscope, a Nikon DS-Fi2 camera paired with a Nikon Eclipse Ni microscope, a Samsung Galaxy J7 Prime smartphone, an iPhone 12 or 13 fitted with a 10× eyepiece of a Nikon Eclipse Ni microscope, or an Olympus BX53 setup. This results in varying resolutions, lighting conditions, and imaging parameters. Some images are out of focus, some suffer from noise, and some display motion blur captured via motorized stage microscopes. The broad spectrum of image quality and visual characteristics yields diverse image features, which is designed to improve the robustness of object detection models. The training and test subsets will be randomly split. We plan to release 1,000 and 250 images per class for training and testing, respectively, on the day of the dataset's official launch. This is the largest dataset of its kind currently available. Annotations will be provided in the form of bounding boxes.

搜集汇总

背景与挑战

背景概述

Chula-ParasiteEgg-11是一个寄生虫卵显微图像数据集，包含11种类型、来自多种设备的图像(分辨率/光照条件各异)，具有图像质量多样化的特点(含失焦/噪点/模糊样本)，旨在提升检测模型鲁棒性，采用边界框标注，是同类中规模最大的数据集。