Turn an FTIR spectrum image into usable numerical data.
Upload a spectrum image or published figure, identify the plotting area, trace the curve, calibrate the axes and export the result for further analysis.
Recommended: use a spectrum image at least 800 px wide.
A clean, untilted screenshot works best; phone photos usually need manual point-by-point tracing.
Three extraction methods are available: automatic, color-based, and manual point-by-point tracing.
Preview the traced curve, then download the CSV and use it for searching — a direct image search can take hours.
The image-to-spectrum extraction tool is another separate path: spectrum images and the extracted curve/CSV data from that tool are used to improve the curve-detection model, but they are not used as reference-library records for retrieval. If you prefer not to have your image or extracted curve/CSV data used for model training, upload your raw instrument file through the standard search flow instead. Raw instrument files submitted through the standard flow remain outside model training and public-library contribution. This tool is open source — the code is available on GitHub for self-hosting and custom configuration.
Upload Spectrum Image
Upload a clear image with low compression noise and strong axis/curve contrast. Uploads through this tool are used to improve the curve-detection model — see the note above.
Curve Preview
Peak-Based Wavenumber Correction
The table below identifies one representative peak in each high, middle, and low wavenumber region. Enter the corresponding correct peak positions, then apply the correction to remap the whole spectrum.
| Band | Detected Peak Candidates (cm⁻¹) | Correct Peak (cm⁻¹) |
|---|
Each dropdown lists automatically marked peaks from that spectral region. All three regions are required, and the corrected peaks must stay in the same spectral order as the selected detected peaks.
Processing...