ChemEq25 :- A Novel Multiclass Dataset for Detecting Chemical Laboratory Apparatus Using Image Processing and Deep Learning Approach

The "ChemEq25" dataset comprises 4,599 annotated images of chemical laboratory equipment, collected from three locations: Bangladesh Atomic Energy Chemical Lab, Biochemical Laboratory of United International University, and Chemistry Laboratory of Dhaka Imperial College. Designed for machine learning applications, it supports real-time equipment detection. Images were captured under varying lighting and backgrounds to enhance model robustness. Each item was photographed from multiple angles, auto-oriented, and resized to 640x640 pixels, introducing slight stretching. The dataset includes 25 classes of common lab equipment, divided into training (70%), validation (20%), and test (10%) sets.

“ChemEq25”数据集包含4599张带标注的化学实验室设备图像，采集自三个场所：孟加拉国原子能化学实验室、联合国际大学生物化学实验室以及达卡帝国学院化学实验室。本数据集专为机器学习应用打造，可支持实时设备检测任务。为提升模型鲁棒性，图像在多样化光照条件与背景环境下采集。所有设备均从多视角拍摄，并经自动朝向校正后调整至640×640像素，调整过程中存在轻微拉伸。该数据集涵盖25类常见实验室设备，并划分为训练集（70%）、验证集（20%）与测试集（10%）。

搜集汇总

背景与挑战

背景概述

ChemEq25是一个专为图像处理和深度学习设计的化学实验室设备检测数据集，包含4,599张标注图像，覆盖25类常见设备，图像采集自多个真实实验室环境以提升模型鲁棒性。数据集已预处理为640x640像素，并按70%/20%/10%的比例划分为训练、验证和测试集，支持实时检测任务。