AI-supported smart monitoring of pollutants in aquatic ecosystems: Real-Time Automatic Detection through TRF Spectroscopic Analysis

This dataset was generated as part of the project “VAI-TRF: Validation of Artificial Intelligence Time Resolved Fluorescence method for the real-time monitoring and automatic identification of critical water pollutants”, conducted at the Institut Lumière Matière (iLM), Université Claude Bernard Lyon 1.The main objective is to develop an automatic pollutant identification method based on Time-Resolved Fluorescence (TRF) measurements combined with artificial intelligence models. Spectroscopic data (TRF method) The data were acquired using an Agilent Cary Eclipse fluorescence spectrophotometer in Time-Resolved Fluorescence (TRF) mode.Each measurement reports the fluorescence intensity for a pair of excitation and emission wavelengths. The excitation wavelengths range from 230 to 520 nm,and the emission wavelengths range from 520 to 610 nm,with an intensity value recorded for each (λ_exc, λ_em) pair. The files are provided as:• Raw CSV files exported directly from the instrument (wide format: repeated Wavelength/Intensity pairs for each λ_em)• 2D/3D images (PNG/JPG) generated from intensity matrices, used as input for AI models Dataset organization The dataset is organized into six pollutant / ligand classes:D – DTPMPO – DOTAN – NTAP – DTPAM – SPO (without pollutant)G – Glyphosate Each folder contains:• TRF spectrum CSV files• Image files derived from CSV data (intensity maps) Usage This dataset can be used for:• Automatic pollutant identification using AI• Training and comparison of classification models (CNN, ResNet, SVM, etc.)• Analysis of spectral signatures of terbium–pollutant complexes Python scripts can be used to:• Convert CSV files into long format (λ_exc, λ_em, intensity)• Generate images from intensity matrices• Train and evaluate AI models本数据集由在光物质研究所（Institut Lumière Matière, iLM）、克洛德·贝尔纳里昂第一大学开展的“VAI-TRF：面向关键水污染物实时监测与自动识别的人工智能时间分辨荧光（Time-Resolved Fluorescence, TRF）方法验证”项目生成。本项目的核心目标为开发一种基于时间分辨荧光测量结合人工智能模型的污染物自动识别方法。 光谱数据（时间分辨荧光法） 本数据集采用安捷伦（Agilent）Cary Eclipse荧光分光光度计在时间分辨荧光（Time-Resolved Fluorescence, TRF）模式下采集得到。每一条测量数据均记录了一组激发波长与发射波长对应的荧光强度。 激发波长范围为230 nm至520 nm，发射波长范围为520 nm至610 nm，且针对每一组（λ_exc, λ_em）对均记录了对应的强度值。 数据集文件包含以下两类： • 仪器直接导出的原始CSV文件（宽表格式：针对每个发射波长λ_em重复存储波长/强度对） • 由强度矩阵生成的2D/3D图像（PNG/JPG格式），可作为人工智能模型的输入数据 数据集组织架构 本数据集共包含六类污染物/配体：D-DTPMP、O-DOTA、N-NTA、P-DTPA、M-DTPAM以及SPO（无污染物组），以及G-草甘膦（Glyphosate）。 每个类别文件夹均包含以下内容： • 时间分辨荧光光谱CSV文件 • 由CSV数据生成的图像文件（强度分布图） 数据集用途 本数据集可应用于以下场景： • 基于人工智能的污染物自动识别 • 分类模型（卷积神经网络CNN、残差网络ResNet、支持向量机SVM等）的训练与对比分析 • 铽-污染物复合物的光谱特征分析 可通过Python脚本实现以下操作： • 将CSV文件转换为长格式数据（λ_exc, λ_em, intensity） • 由强度矩阵生成图像 • 训练并评估人工智能模型