orange-spectra

orange-spectra — spectroscopy widgets for Orange Data Mining

English ｜ 中文

Eleven spectroscopy widgets for Orange Data Mining. They share their algorithms and the .speclib library format with SpectraView, a desktop spectroscopy viewer. Fetch spectra from public databases by URL, compare and search them, build reusable reference libraries, decompose mixtures, and draw aquaphotomics aquagrams — all inside Orange's visual workflow canvas.

Widget	What it does
Import Spectrum URL	Paste an IRUG id/URL or a SOPRANO URL (direct JCAMP-DX/CSV links, and spectra embedded in Plotly/Highcharts/Chart.js charts, work too), download and plot the spectrum, output an Orange Table.
Load Spectra Files	Bulk-load chosen files, a whole folder or a .zip (no extraction): JCAMP-DX, CSV, matrix CSV, NetCDF .cdf → one merged Table.
Spectrometer	Turn a camera + diffraction-grating spectrum photo (Theremino-style) into a calibrated intensity-vs-wavelength spectrum: strip profile + pixel→nm calibration.
Merge Spectra	Overlay several spectra sources on one plot and output a single combined Table (each row a spectrum on a shared grid) — like SpectraView's multi-file overlay.
Spectra Similarity	Score similarity between two sets of spectra: correlation, cosine, spectral angle (SAM), and Euclidean distance.
Spectral Library	Build a reference library, save it as .speclib (interoperable with SpectraView), and rank an unknown spectrum against the library.
Mixture Analysis	Decompose a mixed spectrum with non‑negative least squares (NNLS): solve mixture ≈ Σ cᵢ·refᵢ and report coefficients, proportions, and R².
Aquagram	Aquaphotomics: read normalized absorbance at water's 12 characteristic bands (WAMACs) and draw a 12‑axis radar chart (raw / SNV / aquagram normalization).
Peak Finder	Detect peaks, label them on the plot, and output a peak table (position, height, FWHM, prominence, area).
XRF Element ID	Find peaks in an XRF spectrum (keV) and label them with matching element emission lines (Kα/Kβ/Lα/Lβ, 53 elements Na–U).
PLS-DA	Partial least squares discriminant analysis: class-colored score plot, loadings, VIP variable importance, and predictions.

The output Table follows the Orange-Spectroscopy convention (column names = wavelength/wavenumber, one spectrum per row), so it plugs straight into the Spectra viewer widget or into PCA / PLS chemometrics pipelines.

Install

⚠️ Know which Orange you run first — the desktop App and a pip‑installed Orange are separate Python environments. Installing into the wrong one means the widgets won't appear.

Works on Windows, macOS (Apple Silicon & Intel), and Linux — it's pure Python (numpy / scipy / matplotlib / Orange3).

A. Desktop App (the standalone program from orangedatamining.com): Options ▸ Add-ons… ▸ Add more…, type orange-spectra, tick it, OK, and restart. On macOS, download the Mac .dmg from https://orangedatamining.com/download/.

B. pip Orange (started with python -m Orange.canvas):

pip install orange-spectra
python -m Orange.canvas

On macOS use python3 / pip3 (e.g. install Python via Homebrew: brew install python).

If Orange fails to start with ImportError: PyQt5 … not available, it's missing a Qt binding — install one:

pip install PyQt5 PyQtWebEngine

After (re)starting Orange, a Spectra category with the eleven widgets appears in the toolbox.

Quick start

1. Drop in Import Spectrum URL, enter 4119 (IRUG's PB15 phthalocyanine blue Raman spectrum) → Fetch.
2. Fetch a few reference spectra → feed them to the Spectral Library Spectra input → Add input spectra to library → Save… as a .speclib.
3. Feed an unknown spectrum to the library's Query input → the Hits output is the ranked match table.
4. Feed a mixed spectrum to Mixture Analysis Mixture and the references (or the library's Library output) to References → get component proportions and R².

Every widget has an ℹ How to use box and a 📖 Open tutorial button in its top‑left corner, and F1 (or the ? button) opens its online help page.

Supported URL formats

• IRUG detail pages (jqPlot‑embedded data)
• SOPRANO pages (Dygraph‑embedded data)
• JCAMP-DX (AFFN plain‑number format)
• Two‑column CSV/TSV
• Spectra embedded in Plotly / Highcharts / Chart.js interactive charts (generic x/y fallback)

Compressed JCAMP (SQZ/DIF) is not parsed here — open it in SpectraView and export first.

Documentation

Full tutorial with real‑data demos: https://tai-shengyeh.github.io/spectraview/orange.html (English).

Notes

• You only need to install once; updating requires a reinstall (pip install --upgrade orange-spectra) and an Orange restart.
• Source, issues, and the desktop SpectraView app: https://github.com/Tai-ShengYeh/spectraview.

License

MIT.

---

中文說明

orange-spectra — Orange Data Mining 光譜 widgets

English ｜ 中文

十一個給 Orange Data Mining 的光譜學 widgets， 與桌面版光譜檢視程式 SpectraView 共用演算法與 .speclib 光譜庫格式。可用網址從公開資料庫抓光譜、比對與搜尋、建立可重複使用的 參考光譜庫、拆解混合光譜，並繪製 aquaphotomics 雷達圖——全部在 Orange 的視覺化 工作流程畫布中完成。

Widget	功能
Import Spectrum URL	貼上 IRUG 編號／網址或 SOPRANO 網址（也支援 JCAMP-DX/CSV 直接網址，以及 Plotly/Highcharts/Chart.js 互動圖內嵌資料），下載並畫出光譜，輸出成 Orange Table。
Load Spectra Files	批次載入：選檔、整個資料夾或 .zip（免解壓）——JCAMP-DX、CSV、矩陣 CSV、NetCDF .cdf 全部讀成一個合併 Table。
Spectrometer	把相機＋繞射光柵拍到的光譜照片（Theremino 分光儀）變成校準光譜：取水平帶讀強度、像素→波長校準。
Merge Spectra	把多個光譜來源疊在一張圖，輸出成一個合併 Table（每列一條光譜、共同波段）——等同 SpectraView 的多檔疊圖。
Spectra Similarity	兩組光譜間的相似度：correlation / cosine / 光譜角 SAM / Euclidean。
Spectral Library	建立參考光譜庫、存成 .speclib（與 SpectraView 互通）、對庫比對未知譜並輸出排名。
Mixture Analysis	混合光譜的成分分析：以非負最小平方（NNLS）解 mixture ≈ Σ cᵢ·refᵢ，回報係數、比例與 R²。
Aquagram	Aquaphotomics：在水的 12 個特徵吸收帶（WAMACs）取正規化吸光度，畫 12 軸雷達圖（raw / SNV / aquagram 三種正規化）。
Peak Finder	自動尋峰並在圖上標記，輸出峰表（峰位、峰高、FWHM、顯著度、面積）。
XRF Element ID	XRF 能譜（keV）尋峰並比對元素特徵譜線（Kα/Kβ/Lα/Lβ，Na–U 53 元素），圖上直接標元素。
PLS-DA	偏最小平方判別分析：依類別上色的分數圖、loadings、VIP 變數重要性與預測輸出。

輸出的 Table 採 Orange-Spectroscopy 慣例（欄名＝波長/波數、每列一條光譜），可直接接其 Spectra 檢視 widget 或 PCA / PLS 等化學計量學流程。

安裝

⚠️ 先確認你的 Orange 是哪一種——桌面版 App 與 pip 版是不同的 Python 環境，裝錯不會出現 widgets。

可在 Windows、macOS（Apple 晶片與 Intel）、Linux 執行——純 Python （numpy / scipy / matplotlib / Orange3）。

A. 桌面版 App（orangedatamining.com 下載的獨立程式）： Options ▸ Add-ons… ▸ Add more… 輸入 orange-spectra → 打勾 → OK → 重啟。 macOS 到 https://orangedatamining.com/download/ 下載 Mac 版 .dmg。

B. pip 版 Orange（python -m Orange.canvas 啟動）：

pip install orange-spectra
python -m Orange.canvas

macOS 請用 python3 / pip3（可用 Homebrew 裝 Python： brew install python）。

若 Orange 開不了並出現 ImportError: PyQt5 … not available，是少了 Qt 綁定， 補裝：

pip install PyQt5 PyQtWebEngine

重新啟動 Orange，工具箱會出現 Spectra 分類（11 個 widgets）。

快速上手

1. 拖出 Import Spectrum URL，輸入 4119（IRUG 的 PB15 酞菁藍 Raman 譜） → Fetch。
2. 多抓幾條參考譜 → 接 Spectral Library 的 Spectra 輸入 → Add input spectra to library → Save… 存成 .speclib。
3. 未知譜接 Library 的 Query 輸入 → Hits 輸出就是排名表。
4. 混合譜接 Mixture Analysis 的 Mixture、參考譜（或 Library 的 Library 輸出）接 References → 得成分比例與 R²。

每個 widget 左上角都有「ℹ 說明 How to use」盒子與「📖 開啟線上教學」按鈕；選取 widget 按 F1（或 ? 鈕）會開啟線上說明頁。

支援的網址格式

• IRUG 詳情頁（jqPlot 內嵌資料）
• SOPRANO 頁（Dygraph 內嵌資料）
• JCAMP-DX（AFFN 純數字格式）
• 兩欄 CSV/TSV
• Plotly / Highcharts / Chart.js 互動圖內嵌的 x/y 資料（通用後備解析）

壓縮的 JCAMP（SQZ/DIF）這裡不解析——請先用 SpectraView 開啟後匯出。

教學文件

含真實數據 Demo 的完整教學： https://tai-shengyeh.github.io/spectraview/orange.html （English）。

注意

• 安裝一次即可，更新才需重裝（pip install --upgrade orange-spectra） ＋重開 Orange。
• 原始碼、issues、桌面版 SpectraView： https://github.com/Tai-ShengYeh/spectraview。

授權

MIT。