waterfall-plot

Nature-style 3D filled spectrum waterfall plot — a reusable template that turns a folder of spectra into a publication-ready figure.

Nature 投稿风格的 3D 填充光谱瀑布图模板：给定一个光谱文件夹，一键生成可见光渐变的三维瀑布图。

What it is

waterfall-plot reads a folder of spectra (two columns: wavelength + intensity) and stacks them into a true 3D waterfall plot with filled "curtains". Each spectrum is coloured along the wavelength axis with a visible-light gradient (violet → red by default), producing the classic rainbow look used in spectroscopy papers.

Key properties:

• 3D perspective — three axes: Wavelength / Spectrum / Intensity
• Filled curves — every spectrum is a solid "curtain" from baseline to peak
• Visible-light colour gradient — colour follows wavelength (configurable per band)
• Complete box frame (all 12 edges), unclipped axis labels, Times New Roman
• No hard-coded data — axis limits, spacing and colour mapping adapt to the input
• Outputs 300 + 600 dpi PNG

Repository layout

waterfall-plot/
├── SKILL.md                        # DeepSeek Harness skill documentation
├── scripts/
│   └── make_waterfall.py           # plotting script (generic template)
├── data/
│   ├── spectrum_001.txt … spectrum_010.txt   # 10 spectra, 400–700 nm
│   └── complex_spectra.csv                   # 30-min time-resolved spectrum
└── example/
    └── waterfall_example.png                 # example output

Quick start

pip install numpy matplotlib
python waterfall-plot/scripts/make_waterfall.py data

This writes data/spectra_waterfall_600dpi.png and data/spectra_waterfall_300dpi.png.

Options

Flag	Default	Description
--cmap	visible	visible (rainbow by wavelength), any matplotlib colormap (viridis, plasma, turbo, …), or a #hex solid colour
--spacing	auto	gap between spectra; auto scales to the peak intensity, or a fixed float
--ratio	auto	box proportions x,y,z — e.g. 3,5.5,1.6 stretches the Spectrum axis
--x-label	Wavelength (nm)	x-axis label
--y-label	Spectrum	stacking-axis label
--z-label	Intensity (a.u.)	intensity-axis label
--font	Times New Roman	font family (e.g. Arial)
--figsize	3.9,5.2	figure size in inches width,height
--elev / --azim	24 / -62	3D view angle
--alpha	0.92	fill opacity
--dpi	600	resolution (always also emits a 300 dpi companion)
--out	input folder	output directory
--prefix	spectra_waterfall	output filename prefix

All defaults are also editable in the CONFIG block at the top of make_waterfall.py.

Input data format

Two-column text or CSV: first column = wavelength, second column = intensity. Header and comment (#) lines are skipped automatically. File discovery order: spectrum_*.txt → *.txt → *.csv.

Negative baseline noise is clipped to 0 (a common handling for optical spectra).

Sample data

• data/spectrum_001.txt … spectrum_010.txt — 10 individual spectra, 400–700 nm, 500 points each.
• data/complex_spectra.csv — time-resolved spectrum, one column per minute (30 minutes).

Use as a DeepSeek Harness skill

SKILL.md registers this as a model-invocable skill. Copy the waterfall-plot folder into your DSH skills root (e.g. %APPDATA%\dsh-desktop\harness\skills\), then simply give the assistant a data folder and ask for a "瀑布图 / waterfall".

License

MIT