VIPP — Visual Image Processing Platform

See every decision in your analysis

Visual workflows for reproducible bioimage analysis.

Build bioimage-analysis workflows as connected graphs, inspect intermediate images and tables, then save the workflow for review, adaptation, or batch processing.

Install VIPP Start with a guided example Find a workflow

A complete label-cleanup graph shown in context. For day-to-day authoring, enlarge or undock VIPP so the graph remains the primary work surface.

Alpha release: validate before interpreting

This manual describes the public napari-vipp 0.15.0a1 alpha. Download only from the canonical GitHub release or PyPI page. Verify every intentionally unsigned Windows or macOS installer against its official checksum before opening it. Do not treat unperformed external fresh-machine or novice checks as passes. Interfaces, workflow files, and parameter defaults may change between alpha releases. Treat visual inspection, reference data, and domain review as part of the analysis—not as optional cleanup after it.

Upgrading to 0.15

0.15.0a1 writes workflow schema 6. Valid schema-3, schema-4, and schema-5 workflows migrate explicitly; cached results and source pixels are not saved in workflow JSON, and generated Python is pinned to its creator version. Read the 0.15.0a1 release notes before upgrading and revalidate calculated results afterward.

Batch configs are version 6; manifests remain version 5. Earlier supported records may lack per-sample overrides, SourceItems, or batch execution-profile intent. Review selected items, readers, axes, calibration, migrated values, and effective Run/Bypass and individual existing-file choices before consequential use.

Choose your path

Install VIPPVerify and run the unsigned Windows or macOS installer, or follow a pinned manual route. New to VIPPTour a finished graph and build a small segmentation workflow. I have an analysis taskFollow recipes for segmentation, measurements, networks, colocalization, restoration, or batch runs. I want to use my GPUInstall the standard CUDA choice, choose a compute policy, benchmark safely, and verify what actually ran. I need defensible resultsChoose dimensionality, tune on representative data, validate, and record what must be reported. I know what I needSearch all nodes, bundled samples, example workflows, settings, formats, and compatibility notes.

What VIPP records—and what it does not

VIPP workflow JSON records the graph, node parameters, connections, layout, and selected workflow state. Where available, image state carries axes, scale, units, channel information, and operation history through compatible nodes. This supports inspection and repeat execution, but it does not by itself guarantee scientific reproducibility: input identity, software environment, batch bindings, reference annotations, exclusions, and validation evidence must also be retained.

flowchart LR
  A["Load representative data"] --> B["Build and tune graph"]
  B --> C["Inspect intermediate outputs"]
  C --> D["Validate against references"]
  D --> E["Freeze workflow and environment"]
  E --> F["Run batch and audit outputs"]

Quick links

Goal	Go to
Open a working example in five minutes	Tour a finished workflow
Switch from synthetic data to your images	Use your own images
Understand images, masks, labels, and tables	Data types
Diagnose a workflow that suddenly gives different counts	Common problems
Ask a question or report a reproducible problem	Support routes
Prepare methods and provenance for a paper	Report a VIPP analysis
Choose and verify CPU/GPU execution	CPU and GPU compute
Review everything changed in the public alpha	0.15.0a1 release notes
Contribute a node or documentation fix	Contributor guide

The application is developed in the napari-vipp repository. This site is the quick, searchable manual; slower teaching material can live in a separate course or book.