The job this page compares
Not "which tool is best". One job: three to ten people label multi-channel fluorescence images for a model, and someone has to trust the labels before training. That job has four parts, and tools differ on each.
- Seeing the signal. Each channel needs its own contrast. A dim far-red probe and a bright DAPI stain never share good settings.
- Working on the same images. Either the data lives on a server and everyone opens it in a browser, or each person has files on their own computer.
- Checking the work. A second person reviews, and the accepted set is frozen so the training data does not move.
- Knowing who did what. When a label is wrong, you can find when it changed and who changed it.

The comparison
Every cell about another tool comes from that tool's own documentation, repository or paper, listed under Sources. "Not in the docs we read" means exactly that. It does not mean the tool cannot do it.
| QuPath | napari | Cytomine | WEBKNOSSOS | DataTorch | |
|---|---|---|---|---|---|
| Runs | on your computer | on your computer (Python package or standalone installer) | web platform on your own server | in the browser | in the browser, cloud or your own environment |
| Per-channel contrast | yes, Brightness/Contrast dialog, settings can be saved by name | a multichannel image opens as one image layer per channel | a panel with selectors for channel, z slice and time point, per its 2016 paper. Contrast controls are not in the docs we read | lists Multi-Channel among its image formats | yes, a lookup table per channel |
| Channel names from the file | yes, and you can rename them without losing the originals | not in the docs we read | not in the docs we read | not in the docs we read | yes, from OME-TIFF metadata |
| Several people, same data | a project is a folder of data files and image paths. Sharing is not in the docs we read | not in the docs we read | yes, built for "collaborative analysis" | yes, "share datasets and annotations with collaborating scientists" | yes, in the browser. Collaborators are part of the paid plans |
| Review and sign-off | not in the docs we read | not in the docs we read | proofreading tools and a reviewed layer, per its 2016 paper | proofreading tools for large segmentations, user and task management | review file by file, then sign off the dataset as a locked version |
| History of changes | not in the docs we read | not in the docs we read | not in the docs we read | not in the docs we read | changes are attributed, with an audit trail |
| Z and time stacks | yes, z-stacks and time series | yes, a multi-dimensional viewer | lists 2D+c+z+t | yes, built for large 3D datasets | no, 2D only |
| Whole-slide pyramids | yes | not in the docs we read | yes, "very large bio-medical images" | not in the docs we read | no |
| Cost | free, open source (GPLv3) | free, open source (BSD-3-Clause) | open source (Apache 2.0), you run it | open source (AGPLv3), paid hosting from its makers | Free plan for one builder, Team from $99 per builder a month billed annually, Enterprise per account |
General annotation tools (CVAT, Label Studio, Roboflow) are compared on a different job in Keypoint annotation tools compared.
QuPath
"Open source software for bioimage analysis", free under the GPLv3, and widely used for whole-slide and multiplexed analysis. It reads multidimensional data, z-stacks, time series and multiplexed images among them. Channel names show in its Brightness/Contrast dialog, can be edited by double-clicking, and "the original names are not lost". Brightness and contrast settings "can be saved and re-loaded" by name. A project is a folder that holds "QuPath's associated data files only - and the paths to where your images are located". The project documentation we read does not describe sharing a project between people or reviewing someone else's work.
Choose it if one analyst does the labeling and the analysis, or your images are whole slides.
napari
"A fast, interactive, multi-dimensional image viewer for Python", free under the BSD-3-Clause license. It installs with pip or conda on macOS, Linux and Windows, or from a standalone installer. A multichannel image opens as one image layer per channel, and points, shapes and label layers sit on top. It is a viewer, so where the files live and who checks whose labels is yours to arrange.
Choose it if your team writes Python and each person can manage their own files.
Cytomine
"An open-source platform for collaborative analysis of large-scale imaging data", under Apache 2.0. Its 2016 paper describes proofreading tools with which experts "accept or reject annotations" made by other users or by algorithms, and says "reviewed annotations are stored as novel entities in the database". The same paper shows a "multidimensional image panel with selectors for channel, slice in a z-stack, and time point". The current user guide pages we could open do not describe review or channel contrast, so check the version you install. You run it on your own server.
Choose it if your images are whole slides and you have someone to run a server.
WEBKNOSSOS
"An open-source tool for annotating and exploring large 3D image datasets", under AGPLv3, with a "fully browser-based user experience". Its repository lists Multi-Channel among supported image formats, sharing of datasets and annotations with collaborators, "user and task management for high-throughput crowdsourcing", and "proofreading tools for working with large (over)-segmentations". Its makers, scalable minds, offer commercial hosting. We could not reach its documentation or pricing pages when writing this, so nothing here goes beyond the repository.
Choose it if your data is 3D volumes and many people trace or proofread them.
DataTorch
A browser annotator. Upload a multi-channel TIFF and each page of the file becomes a channel you can show, hide and window on its own, with a lookup table per channel. Channel names stored in OME-TIFF metadata are shown. Uncompressed 16-bit data is read at full depth and windowed down to 8-bit for display.
Review is file by file: each file in a task has its own status, and a reviewer works through the labels one file at a time. Sign-off is on the whole dataset: one person with the sign-off permission cuts a named version of every file and label in it. The dataset is then read-only until someone with the unlock permission opens it for amendment, and that unlock is recorded in the version history. Changes are attributed to the person who made them, with an audit trail.
It runs in the cloud, or deployed in your own environment on the Enterprise plan. The Free plan covers one builder with unlimited public projects and one private project. Collaborators come with the Team plan, where reviewers and annotators do not need paid seats.
What it does not do today: Z and time stack navigation, compressed or tiled OME-TIFF planes, whole-slide pyramids, Bio-Formats conversion.
Choose it if your images are 2D multi-channel fields, several people label them, and someone has to sign off on the set before training.
Step by step on a real dataset: the multi-channel TIFF tutorial.
Sources
Other tools read on November 13, 2026. DataTorch pricing read on November 16, 2026.
- QuPath, multiplexed analysis: https://qupath.readthedocs.io/en/stable/docs/tutorials/multiplex_analysis.html
- QuPath, projects: https://qupath.readthedocs.io/en/stable/docs/tutorials/projects.html
- QuPath, supported image formats: https://qupath.readthedocs.io/en/stable/docs/intro/formats.html
- QuPath repository (license): https://github.com/qupath/qupath
- napari repository (description, license): https://github.com/napari/napari
- napari, installation: https://napari.org/stable/getting_started/installation.html
- Cytomine documentation (image types): https://doc.uliege.cytomine.org/
- Cytomine repository (description, license): https://github.com/cytomine/cytomine
- Cytomine paper: Marée et al., "Collaborative analysis of multi-gigapixel imaging data using Cytomine", Bioinformatics 32(9), 1395-1401 (2016): https://pmc.ncbi.nlm.nih.gov/articles/PMC4848407
- WEBKNOSSOS repository: https://github.com/scalableminds/webknossos
- DataTorch pricing: https://datatorch.io/pricing
Label multi-channel images with your team
Upload a multi-channel TIFF, set a lookup table per channel, and review each other's labels in the browser.
Frequently asked questions
Is there a free microscopy annotation tool for teams?
QuPath (GPLv3) and napari (BSD-3-Clause) are free and open source, and run on each person's computer. Cytomine (Apache 2.0) and WEBKNOSSOS (AGPLv3) are open-source web platforms that you or a host run. DataTorch's Free plan covers one builder, with unlimited public projects and one private project with one dataset. Collaborators and more private projects come with the paid Team plan.
Can several people annotate the same microscopy images at once?
In a web platform, yes, because the images and labels live on a server. In a desktop tool each person works on their own computer, and sharing the files is up to the team.
Which tool should I use for 3D stacks?
Not DataTorch today. It shows 2D multi-channel images and does not navigate Z or time. napari is a viewer for multi-dimensional images on the desktop, and WEBKNOSSOS is a browser tool for large 3D datasets.
Open the hosted copy on DataTorch.
Browse the files in your browser, then download the dataset with a free account.
