changelog#

v0.0.11 (2026-02-25)#

Shared#

  • canvas resize unlocked — drag-resize on all widgets is no longer capped at 600–800px; resize as wide as your screen allows

v0.0.10 (2026-02-25)#

Shared#

  • Arina GPU deps now required — h5py, hdf5plugin, numba, and pyobjc-framework-Metal (macOS) are installed automatically; no manual setup needed for GPU-accelerated Arina loading

  • graceful MemoryError when Metal buffer allocation fails (e.g. det_bin=1 on a dataset larger than RAM) — suggests using a larger det_bin instead of crashing with a cryptic error

v0.0.9a1 (2026-02-23)#

Shared#

  • IO module — unified file loader for all common EM formats: IO.file("image.dm4") returns an IOResult with data, pixel_size, units, and title extracted from file metadata; IO.folder("scans/") stacks a folder of files with optional file_type filter and recursive flag; IO.folder([folder1, folder2]) merges multiple folders into one stack; IO.file([path1, path2]) reads and stacks multiple files/folders; supports PNG, TIFF, EMD, NPY/NPZ natively and DM3/DM4, MRC, SER via rsciio; IO.supported_formats() lists available extensions; folder auto-detects file type when file_type is omitted

  • IO.arina_file() — GPU-accelerated arina 4D-STEM loader: IO.arina_file("master.h5", det_bin=2) with auto-detected GPU backend (MPS now, CUDA/Intel coming soon), det_bin="auto" (RAM-based), scan_bin (navigation binning), and hot_pixel_filter (on by default). IO.arina() renamed to IO.arina_file() for consistency with IO.arina_folder(). IO.arina() still works as an alias.

  • IO.arina_file(list) — cherry-pick specific files into 5D: IO.arina_file(["scan_00.h5", "scan_03.h5"], det_bin=4) stacks selected master files into a 5D tensor; raises ValueError on shape mismatch

  • IO.arina_folder() — batch 5D-STEM loading: IO.arina_folder("session/", det_bin=8) finds all *_master.h5 files, loads each with IO.arina_file(), and stacks into a 5D array; supports multi-folder (IO.arina_folder([folder1, folder2])), recursive=True for nested subdirectories, pattern="SnMoS2" for name filtering, and max_files=50 to cap loaded files; incomplete files are auto-skipped with a warning; shape-mismatched files are skipped with a warning

  • per-frame metadata — IO.arina_file(list) and IO.arina_folder() extract full HDF5 metadata from each source file into IOResult.frame_metadata (schema-agnostic — stores every scalar dataset as-is)

  • IOResult.describe() — print a per-frame metadata table; describe() (default diff=True) shows only columns where values differ across frames; describe(diff=False) shows all; describe(keys=["count_time"]) for custom columns

  • all widgets now accept IOResult directly — Show2D(IO.file("image.dm4")) passes data, pixel_size, title, and labels automatically

Show4DSTEM#

  • 5D IOResult from IO.arina_folder() now passes frame labels and title through automatically — Show4DSTEM(result, frame_dim_label="Scan") shows frame slider with file names

  • free() — convenience method to release GPU memory: deletes MPS tensor, runs garbage collection, and flushes MPS allocator cache back to system

Show2D, Show3D, Show4D, Show4DSTEM, Mark2D#

  • FFT now renders correctly for non-power-of-2 images (e.g. 96x96 diffraction patterns from arina loader) — pre-pads to next power of 2 before FFT to preserve all frequency bins

  • ROI FFT Hann window: a 2D Hann window is automatically applied before FFT when viewing an ROI region, eliminating spectral leakage streaks from rectangular crop boundaries — Show2D exposes a Win: toggle (default on) to disable windowing; other widgets apply it unconditionally

v0.0.7 (2026-02-21)#

New widgets#

  • Show1D — interactive 1D viewer for spectra, profiles, and time series with multi-trace overlay, calibrated axes, log scale, and figure export

Show2D, Show3D, Mark2D#

  • breaking: image_width_px renamed to canvas_size — pass canvas_size=800 to set the canvas display width in pixels (0 = auto)

Show1D#

  • auto-contrast with percentile clipping (default 2–98%)

  • selectable peak markers with snap-to-local-max, Delete to remove

  • grid density slider (5–50 lines)

  • axis range lock: drag on X or Y axis to lock, double-click to unlock

  • range-scoped statistics: mean/min/max/std and integral within locked range

  • peak FWHM measurement with Gaussian fit overlay

  • CSV export (range-only or full) from Export dropdown

  • figure export (PDF + PNG) from Export dropdown

  • legend click to focus/unfocus traces

Show2D#

  • stack reduction modes (first, index, mean, max, sum) for collapsing stacks to 2D

  • real-time ROI FFT updates during drag (no longer deferred to mouseup)

  • layout: stats bar and controls now match canvas width

Show3D#

  • diff mode: diff_mode="previous" or "first" for frame-to-frame or cumulative change

  • profile_all_frames(): extract the same line profile from every frame

  • multi-ROI: place, resize, duplicate, and delete multiple ROIs with per-ROI color and stats

  • one-click export bundle (.zip with PNG + ROI timeseries CSV + state JSON)

Edit2D#

  • breaking: mask mode now uses rectangle tool only (brush, ellipse, threshold removed for stability)

  • undo/redo for mask operations (Ctrl+Z / Ctrl+Shift+Z) with 50-step history

  • brush_size trait removed from state_dict

  • per-image independent editing (shared=False) with Link toggle

  • ArrowUp/ArrowDown nudge, Shift+drag aspect lock, histogram range labels

Show2D, Show3D, Mark2D, Show4D, Show4DSTEM, ShowComplex2D, Show3DVolume#

  • ROI FFT: FFT shows cropped ROI region with real-time updates during drag

  • d-spacing click: click FFT panel to measure d-spacing with sub-pixel Bragg spot snap

ShowComplex2D#

  • single-mode ROI (circle, square, rectangle) with drag and resize

Show3DVolume#

  • single-mode ROI on XY slice (circle, square, rectangle)

Show2D, Show3D, Show4D, Show4DSTEM, ShowComplex2D, Show3DVolume#

  • Shift+drag rectangle ROI corner locks aspect ratio

Show3D, Show4D, Show4DSTEM#

  • quick-view presets: save/recall 3 display configurations (1/2/3, Shift+1/2/3)

Bin#

  • Bin widget for calibration-aware 4D-STEM binning with BF/ADF QC previews

  • export: save_image() (single panel or grid PNG/PDF), save_zip() (all panels + metadata), save_gif() (original vs binned comparison)

Batch preprocessing#

  • preset-driven folder batch runner: python -m quantem.widget.bin_batch

  • supports .npy and .npz, plus streamed 5D .npy processing for time_axis=0

  • torch device can be selected via preset or --device

v0.0.6 (2026-02-19)#

Show4DSTEM#

  • save_image() for programmatic export to PNG/PDF with optional metadata sidecar JSON

  • export views: diffraction, virtual, fft, all

  • save_image() supports temporary overrides (position=(row, col), frame_idx) with automatic state restoration

  • exported images now match interactive display settings (colormaps, scale modes, percentile clipping)

  • state_dict() now includes scan position (pos_row, pos_col) for exact state restore

  • ArrowUp / ArrowDown scan-row navigation (with Shift step), Esc to release focus

Show4D#

  • ArrowUp / ArrowDown row navigation (with Shift step), Esc to release focus

All profile widgets (Show2D, Show3D, Show4D, Show4DSTEM, Mark2D)#

  • line-profile endpoints are now draggable after placement

  • dragging the line body translates both endpoints together (preserves line shape)

  • hover cursor changes near profile endpoints/line

v0.0.5 (2026-02-18)#

All profile widgets (Show2D, Show3D, Show4D, Show4DSTEM, Mark2D)#

  • set_profile now takes two (row, col) tuples: set_profile((row0, col0), (row1, col1))

Show4DSTEM#

  • 5D time-series/tilt-series support: accepts (n_frames, scan_rows, scan_cols, det_rows, det_cols) arrays with frame slider, play/pause controls, and frame_dim_label (e.g. "Tilt", "Time", "Focus")

  • frame playback: fps slider, loop, bounce, reverse, transport buttons, [ / ] keyboard shortcuts

  • grab-and-drag ROI: clicking inside the ROI drags with an offset instead of teleporting the center

  • theme-aware ROI colors: darker green overlays in light theme for better visibility

  • fixed resize handle hit area: was ~70px due to pixel mismatch, now correctly sized

v0.0.4 (2026-02-16)#

Show2D#

  • ROI with inner/outer resize handles and cursor feedback

  • line profile with hover value readout and resizable height

  • live FFT filtering with mask painting

  • colorbar overlay, export dropdown, clipboard copy

  • auto-contrast with percentile clipping

  • gallery mode with keyboard navigation

Show3D#

  • ROI and lens controls moved to toggle panels

  • line profile with hover value readout and resizable height

  • live FFT filtering with mask painting

  • FFT panel aligned with main canvas

  • colorbar overlay, export dropdown (figure, PNG, GIF, ZIP), clipboard copy

  • auto-contrast with percentile clipping

Show3DVolume#

  • three orthogonal slice panels (XY, XZ, YZ) with synchronized cursors

  • export figure with all three slices

  • colorbar overlay, GIF/ZIP export

Show4DSTEM#

  • virtual imaging with BF, ABF, ADF, custom ROI presets

  • diffraction pattern viewer with annular ROI

  • export dropdown and clipboard copy for both panels

  • colorbar overlay

Show4D#

  • dual navigation/signal panel layout

  • ROI masking on navigation image

  • path animation with GIF/ZIP export

  • export dropdown and clipboard copy

Align2D#

  • two-image overlay with opacity blending

  • FFT-based auto-alignment via phase correlation

  • difference view mode

  • export figure for both panels

Mark2D#

  • interactive point picker with click-to-place

  • ROI support (rectangle, circle, annulus)

  • snap-to-peak for precise atomic column positioning

  • undo, colorbar overlay, export figure with markers

Shared#

  • light/dark theme detection across all widgets

  • colormap LUTs (inferno, viridis, plasma, magma, hot, gray)

  • WebGPU FFT with CPU fallback

  • HiDPI scale bar with unit conversion

  • publication figure export via exportFigure

  • state persistence (state_dict, save, load_state_dict, state param)

  • set_image for replacing data without recreating widget

  • NumPy, PyTorch, CuPy array support

  • quantem Dataset metadata auto-extraction

  • (row, col) coordinate convention

v0.0.3 (2025-12-01)#

  • initial release with Show2D, Show3D, Show3DVolume, Show4DSTEM

  • demo notebooks

  • sphinx docs with pydata theme