Graph display settings¶
These controls change graph-card presentation, not image data or analysis results. They are separate from the main napari viewer's display controls.
In 0.15.0a1, open Preview. In nightly builds after 0.15.0a1, the button is Display settings, shortened to Display in narrower windows.
Menu labels¶
| 0.15.0a1 label | Nightly label |
|---|---|
| Mode | Thumbnail view |
| Contrast | Contrast method |
| Range | Contrast based on |
| Colormap | Colour map |
| Detail | Thumbnail resolution |
| Port labels | Input/output labels |
The options below use the nightly wording. Slice / MIP / Off in 0.15.0a1 correspond to Current slice / Maximum-intensity projection (MIP) / Hidden. Percentile / Min-max / Raw become Percentile (0.5–99.9%) / Minimum–maximum / Raw; these are label changes, not new calculations.
Nightly builds disable Entire stack when no available graph preview contains a stack or series. The view-dependent label describes what is displayed; this does not change the saved preference for use when a stack becomes available.
Choose a useful view¶
- Use Current slice for interactive inspection, or Current image for 2D.
- Use Maximum-intensity projection (MIP) for an overview, not to verify 3D connectivity or depth.
- Use Hidden when thumbnails are unnecessary or expensive.
- Use Entire stack contrast for stable brightness while moving through slices.
- Use current-slice/view contrast for quicker local inspection. Its limits can differ between slices; do not read equal brightness as equal intensity.
Resolution and exact statistics¶
| Resolution | Backing pixels |
|---|---|
| Low | 90 × 55 |
| Standard | 180 × 110 |
| High | 360 × 220 |
| Very high | 720 × 440 |
Resolution changes the retained thumbnail image, not node calculations or scientific output. The graph card stays the same size; Very high retains four times the backing pixels of High, not necessarily more screen pixels.
- Entire stack contrast keeps its cached exact limits when resolution changes.
- Current slice contrast uses the selected resolution's spatial sample, so its display limits may change slightly.
- These are local presentation preferences, not workflow parameters or scientific provenance.
Where to see contrast activity¶
Select a node and look for Thumbnail contrast in its inspector. It shows Calculating…, CPU · NumPy, GPU · CuPy, amber CPU fallback, or red Error. This is separate from the graph card's scientific-compute badge.
Hover the row or thumbnail for scope, resolution, algorithm, byte count, elapsed time, backend choice, and failure details. Keyboard What's This help and screen readers receive the same information.
The shared toolbar reports the active node, backend, and statistics phase. Cancel takes effect at the next safe boundary; a GPU kernel or exact NumPy percentile call may need to finish first. No partial contrast limit is published.
How the limits are calculated¶
For Stack Percentile contrast, native uint8 and uint16 results use an exact
dtype-aware histogram. CPU and CuPy produce the same display limits, including
the NumPy-linear 0.5th/99.9th-percentile result. Min-max uses an exact native CPU
reduction rather than constructing a histogram. Float and other-dtype
percentiles use the exact NumPy-compatible CPU path in this release. Raw integer
contrast, masks, labels, tables, and other scan-free previews do not launch an
unnecessary statistics scan.
CPU and GPU choices¶
- Presentation Auto chooses using the full output's dtype and size, not thumbnail resolution.
- Presentation CPU never initializes CUDA.
- Presentation Prefer GPU attempts eligible CuPy histograms and visibly falls back to CPU when safe.
The main compute policy takes precedence: CPU forces presentation CPU; Prefer GPU biases presentation Auto toward GPU; Auto and Custom leave it adaptive. Automatic selection is a heuristic, not a guarantee of the fastest path on every dataset or machine.
Automatic-selection and background-work details
Before the first successful thumbnail GPU calculation, Auto uses a
384-MiB crossover for uint8 and 512 MiB for uint16. Both become 32 MiB
while that path is warm. Data distribution, hardware, CUDA startup,
residency, and competing work can change which path is faster.
Small CPU-only batches finish immediately without occupying the toolbar progress strip: at most 1 MiB total, eight requests, and eight aggregate scalar/channel lanes. Larger work, high-channel data, and GPU paths use a cancellable background worker. Inspector status is recorded in either case.
CPU integer histogram and min-max calculations advance and cancel between bounded chunks. Scheduling changes neither the population nor the contrast algorithm.
Port-label layout¶
When needed / Always / Never in nightly builds correspond to Ambiguous only / Show all / Hide all in 0.15.0a1.
Long port names are shortened on the card and retain their full text in a tooltip. Changing the label mode can make an already tightly packed layout overlap; VIPP reports the number of overlapping card pairs in its message strip. Use Auto Arrange to make label-aware space, or move the affected cards manually. Label visibility is a graph-display choice and never changes connections or processed data.
Related settings¶
Link napari/VIPP sliders controls dimension synchronization. Save thumbnail visibility in workflows saves per-node visibility choices. Both remain in the gear menu; neither changes pixels.
For full-resolution inspection and exported figures, see inspect and compare outputs.