Give everyone a clear role.
Keep collaborators, decisions and questions in the same study. Know what is ready and what needs attention.
Lab Network
Your labs.
One shared study.
Agree the protocol, keep the discussion together, and run research rounds on each lab’s local images. Review the evidence side by side, with a clear record of what changed.
Training images stay on each lab’s computer. Model updates and agreed research summaries are exchanged.
Already invited? Join a study
How a study comes together
Training labImages stay local
Training labImages stay local
Your shared studyProtocol · discussion · research rounds
Held-out labEvaluates without taking part in training
Keep training close to the data.
Each lab prepares its own images and approves its participation. Local clients carry out the agreed research rounds.
Read the evidence, site by site.
Compare the shared candidate with local baselines. Keep held-out results, sample counts and the study history together.
A useful home for the work
All the context.
A clear next step.
Start with a question your collaborators can work on together. Your study keeps the protocol, conversation and deliberately shared results within reach.
Open the study workspaceWhat belongs in your study
01
The agreed protocol
A shared question and written plan everyone can return to.
02
The conversation around it
Questions, decisions and context, with their authors attached.
03
The results you choose to share
Preview a summary of reviewed observations before publishing it to the study.
From the running workspace. Saved, reviewed summaries and the conversation around them. This walkthrough uses generated images and simulated participants; it shows the interface, not real-lab performance.
Inspect the full screenshot →Connected federated learning
Local images.
Learning you can examine.
Test whether learning across your participating labs helps on a lab the model has not seen.
The client runs on each lab’s computer. Training advances while the required clients stay connected.
Explore the research method →- Prepare and approve locally
Each lab prepares its annotated images, chooses its role and records separate data-owner consent.
- Freeze the plan. Let the rounds run.
The owner locks the roster and settings. Connected clients train locally and exchange the required updates.
- Compare, review and keep the record
Local baselines and held-out evaluation lead to a downloadable report and research candidate.
Lab Network · Shared research
Room for your collaborators.
A plan for the work.
One coordinating account brings the people, reviewed summaries and connected research rounds together.
- 3 active studies, with up to 10 members including the owner and 5 connected labs per study
- A shared protocol, up to 500 retained summaries and 1,000 discussion messages per study, and study exports
- Automatic research rounds with connected local clients
- 100 completed aggregation rounds each UTC month across your studies; up to 20 rounds per run
For the paying account: 1,500 Petri analyses and 1,000 assay analyses per UTC month, 48 photos per batch and 300 saved photos. Invited members keep their individual analysis and storage allowances.
€99/ month
Per coordinating account.
Invited members can join with a free account.
Monthly subscription · Review the order before payment
See how a shared study worksBefore you start
A few useful details.
Start with collaborators you already know and one question you can evaluate together.
Does joining a study share my existing experiments?
No. Your projects and photographs remain in your account. You choose which reviewed observations to summarize, inspect the preview, and publish the aggregate summary yourself. Joining also does not enroll your images in training.
Who needs a paid subscription?
The study owner needs Lab Network to create and maintain active studies. Invited contributors and viewers can join with free accounts. The owner’s included analysis capacity is personal; each invited member keeps their own allowance.
What do I need to run a learning study?
At least two training accounts and one separate held-out account, prepared local datasets, and explicit permission from each data owner. Install the Python client on each participating computer and keep it connected for the required steps. The setup guide walks through this.
What does a completed run give us?
A research report comparing the final candidate with local baselines, per-site aggregate evaluations, and a downloadable candidate for local research predictions. Completion does not establish a real-world accuracy improvement or activate the candidate in DishFlow. Read the method and evidence boundary.
What happens if I leave a study?
Leaving removes your shared summaries and discussion. Hosted learning artifacts are also removed for every frozen run involving your lab, including completed runs. The same applies when you are removed or delete your account. Copies that members have already downloaded cannot be recalled.
What happens if the owner’s subscription ends?
The study and its exports remain readable. New content, invitations and learning pause until the owner restores Lab Network. Members can still withdraw their shared content and participation.