arrayview 0.9.0

Fast multi-dimensional array viewer

arrayview

A viewer for multi-dimensional arrays.

• CLI and Python
• Jupyter / VS Code
• Browser / native
• SSH / tunnels

CLI

uvx arrayview scan.nii.gz
uvx arrayview --window browser scan.npy
uvx arrayview                            # demo

Python

from arrayview import view
view(arr)

MATLAB

Add the matlab/ directory to your MATLAB path, then:

addpath('/path/to/arrayview/matlab')

A = rand(100, 200, 10);
arrayview(A)

Requires arrayview installed in MATLAB's Python environment:

pip install arrayview

Arrays are passed zero-copy via the buffer protocol (in-process Python). arrayview() enables this automatically — just call it before any other py.* call in your MATLAB session.

PyTorch / Deep Learning

from arrayview import view_batch, TrainingMonitor

# Browse a DataLoader batch
view_batch(train_loader)
view_batch(train_loader, overlay='label')

# Live training monitor — updates every N epochs
monitor = TrainingMonitor(every=5, samples=3)
for epoch in range(100):
    for batch in val_loader:
        pred = model(batch['image'])
        monitor.step(input=batch['image'], target=batch['label'],
                     prediction=pred, epoch=epoch)

view_batch() accepts DataLoaders, Datasets, dicts, tuples, or raw tensors. TrainingMonitor opens a compare window and calls handle.update() automatically. PyTorch is not required at import time.

Formats

.npy .npz .nii .nii.gz .zarr .pt .h5 .tif .mat

Once open

Navigation: scroll slices · h/l cycle dims · j/k slices · =/- zoom · drag pan Views: v 3-plane · z mosaic · q qMRI · n compare · = immersive Display: c/C colormaps · d/D dynamic range · f FFT · m complex · p projections · L log Tools: S segmentation · u ruler · s screenshot · ? help

nnInteractive Segmentation

S starts AI-assisted 3D segmentation (requires CUDA). Click/draw to segment, Enter to accept.

[nninteractive]
url = "http://gpu-server:1527"   # skip auto-launch, use running server

Or: ARRAYVIEW_NNINTERACTIVE_URL=http://gpu-server:1527

Config

~/.arrayview/config.toml:

[viewer]
colormaps = ["gray", "viridis", "plasma"]   # colormaps cycled by 'c'

[window]
default = "browser"                         # browser | native | vscode | inline

Full documentation →