Run your AI model
What you'll get: your model run on all your images, with results grouped by folder so you can compare conditions. About 1 minute of work.
A model is Active as soon as training finishes, and ready to run. It runs on any number of images on Biodock's cloud computers.
1. Start the run from your model

Open your project, click the model version on the All models tab, and click Use for analysis. The New analysis page opens. Under Analysis type, click Automated batch: your model is already picked under Pipeline.
You can also start from Analysis → Start new analysis: under Analysis type, choose Automated batch, then under Pipeline pick your project from Your models. If it has more than one active version, choose the version on its card.
Under Input files, click Add files. In Choose images, tick the images or folders to analyse (a folder brings every image inside it), and click Confirm selection.
Under Parameters, check Channel mapping: which channel of your images is which. If your images are the same kind the model learned on, there is nothing to change. All the images in one run need the same number of channels.
Optionally, pick a Results folder to file the analysis in.
In the Job manifest on the right, type a name under Job name. The manifest lists anything still missing in red, and shows what the run will cost (see Credits, storage and limits).
Click Submit.
Every option on this page is explained in Running an analysis.
Folders become groups
Each folder you pick becomes an image group in the results, so you can compare groups straight away, for example Control against Treatment. Before you run an analysis, put each condition in its own folder. The page shows the groups under Results will be arranged into the following groups.

Who will see the results
In a workspace, a new analysis is private: above Submit you see Only you will see the results. If you pick a Results folder that others can see, the line changes to Others will see this analysis. See Sharing and visibility.
A model belongs to the place it was trained in. A model trained in a workspace runs only in that workspace, and one trained in your personal account runs only there.
2. Follow the run
Your run appears at the top of the Analysis list with a progress bar. Click Watch live to follow it on the Job details tab. It goes from Queued to Processing to Succeeded. It is safe to close the tab and come back later.
Bigger and more images take longer; a run usually takes between 6 and 30 minutes.
3. Read your results
When the run has finished, click it in the Analysis list. It opens on the Dashboard tab, with statistics, plots and your images with their outlines. Download your numbers from Export data. See Results dashboard overview.
Keep your list tidy
When a newer version is better, take the old one off the list so nobody runs it by mistake. On the old version, click Deactivate and confirm. Nothing is deleted, and old analyses keep their results. Activate model brings it back. Only project owners can do this.
Next steps
Labeling shortcuts: every keyboard shortcut in the labeler.
Configuring your model: predict more or fewer objects, change what is measured, and set up hierarchies.
Evaluating performance: what the numbers in Training statistics mean.
Last updated