Runs

The Runs view shows all model comparison runs — batch jobs that re-run one or more AI models against a set of images. Use this view to track progress, inspect results, and clean up old runs.

Runs table

The main table lists comparison runs sorted by creation date, newest first. Up to 100 runs are shown.

Column	Description
Checkbox	Select runs for bulk deletion. The header checkbox selects all.
Status	The current state of the run: pending, running, completed, or failed.
Created	When the run was created.
Models	The AI models included in this run, shown as badges.
Priority	The run’s priority level.
Progress	A progress bar showing how many images have been processed out of the total.
Actions	A View button to open the run detail page.

Note

If no runs exist yet, the page shows an empty state with a button to create your first comparison.

Creating a new run

Comparison runs are created from the Home view:

1. Filter and select images on the Home page.
2. Click “Compare selected” in the filter pane to open the comparison page.
3. Choose one or more models to run against the selected images.
4. Set parameters — sleep between entries (how long to wait between processing images) and batch size.
5. Click “Run Comparison” to start the run.

After creation, the run appears in the Runs table. You can navigate back here at any time to check its progress.

Tip

You can navigate to the Runs view while a comparison is still processing. The table updates to show real-time progress.

Viewing run details

Click View on any run to open its detail page. The detail page has two sections:

Run summary

Field	Description
Run ID	A unique identifier for this run.
Models	The models used, shown as badges.
Inference time	Average and median inference time per model, with the number of samples measured.
Priority	The priority level of the run.
Status	Current state of the run.
Created	When the run was created.
Start time	When processing began, or “Immediate” if it started right away.

Results table

The entries table shows each image processed in the run. Entries are sorted by how much the models disagree — the most interesting cases appear first.

Column	Description
Status	A colored dot: gray (pending), blue and pulsing (processing), green (completed), red (failed).
Difference	Whether the models agree. For classification: “All Match” (green) or “All Different” / “X/Y Agree” (red). For object detection: based on which classes were detected. For coordinate models: pixel distance between predicted locations.
Conf. diff	The confidence difference between models: low (green, under 10%), medium (yellow, 10–30%), or high (red, over 30%).
Model columns	One column per model showing its prediction for that image.
Error	Error details if the entry failed.
Actions	A View button to open the image in annotation view with the comparison results overlaid.

Tip

Entries are sorted so the images where models disagree most appear at the top. Focus your review there — agreement means the models are likely both right or both making the same mistake.

Auto-refresh

The detail page polls for updates every 5 seconds while the run is in progress. Auto-refresh stops automatically when the run completes or fails. You can toggle it off manually.

Deleting runs

1. Select runs using the checkboxes in the table.
2. Click “Delete Selected” — the count of selected runs is shown on the button.
3. Confirm the deletion. The button changes to “Confirm Delete”. Click it again to proceed, or click “Cancel” to back out.

Caution

Deleting a run removes it and all its results permanently. This cannot be undone.

What’s next?

Compare model versions A step-by-step guide to evaluating a new model version.

Thresholds Statistics Adjust confidence thresholds based on what you learn from comparisons.

Home Return to image browsing.