general-plot-digitizer

What problem does it solve?

This Skill extracts calibrated numeric X–Y data from experimental spectrum plot images, so you can avoid manually reading points off figures and graphs.

Core Features & Use Cases

• Agent-in-the-loop plot digitization: Uses a vision-language model to inspect axes/bounding boxes/curve obstacles, then a coding pipeline to isolate curves and convert pixels into data coordinates.
• Supports many plot types: Handles Raman, XRD, UV-Vis, IR, NMR, and similar plots with linear/log axes and inverted directions.
• Multi-curve and stacked spectra: Extracts multiple traces from a single figure using per-curve regions, color hints, and optional per-curve normalization.

Quick Start

Run the digitization pipeline on a source plot image using provided metadata by executing: python .agents/skills/general-plot-digitizer/scripts/digitize_pipeline.py examples/source.png --full --metadata examples/metadata.json --all-curves --output-dir ./output.