QCLM¶

QCLM (Quantum Cascade Laser Microscopy) is an open toolkit for hyperspectral infrared imaging. It has two parts:

a desktop app for instrument control, live acquisition and spectral analysis
a backend that runs on your machine: it builds and calibrates spectral cubes, stores your data and runs classification

You install the desktop app like any other application and start the backend with a single Docker command.

macOS Apple Silicon (.dmg) macOS Intel (.dmg) Windows 64-bit (.msi) Linux Debian / Ubuntu (.deb)

Then follow Installation and Backend setup. Getting to a first scan takes about ten minutes.

What you can do with it¶

Control the instrument. Tune the laser by wavenumber, set the pulse rate, coadds and frame rate, and move the X/Y/Z stage. Choose single-wavelength, full hyperspectral or discrete (sparse) wavelength scans.
Watch acquisitions live. Frames stream into the viewer as they are captured, and a spectral profile follows the cursor.
Open existing data. Import sessions saved as ENVI (.hdr), HDF5, NetCDF, NumPy (.npy), TIFF, Zarr or raw binary.
Calibrate and clean. Apply white/dark reference correction, destriping, normalization, band cropping, downsampling and ROI cropping.
Mask and select regions. Draw ROIs and sub-ROIs, threshold, combine and clean masks, and define a measurement ROI.
Reduce dimensionality. Run MNF or PCA transforms, view eigenvalue plots, pick bands by correlation, and extract end-members with the Pure Pixel Index (PPI).
Compute spectral indices. Use the built-in RABD, RABA, ratio and mean-reflectance indices, or define your own and export the maps and plots.
Classify. Run a classification model on a cube and compare models side by side.
Automate. Run batch preprocessing and batch index computation over whole folders, and export or import complete projects.

Project status

QCLM is under active development.

Simulated laser: the backend currently drives a simulated laser and detector, so you can learn the full workflow without hardware. Drivers for real instruments plug in behind the same interface.
Placeholder classifier: the bundled classification model is a simple reference model (a threshold classifier) that shows the inference workflow. It is not trained for any particular sample type.

How it fits together¶

flowchart LR
    A["QCLM desktop app"] -- "REST + live events" --> G["API gateway :8080"]
    subgraph B["Backend (Docker, on your machine)"]
        G --> E[Experiment]
        G --> Q[Acquisition]
        G --> P[Processing]
        G --> I[Inference]
        G --> M[Metadata]
        G --> D[Data storage]
        Q -- frames --> K[(Kafka)]
        K --> P -- cubes --> K --> I
    end

The desktop app only talks to the gateway. By default that is http://localhost:8080, but you can point the app at a backend on another machine in your lab. System architecture explains each service.

QCLM¶

What you can do with it¶

How it fits together¶

Next steps¶