Skip to main content
Once you’ve trained a few models, this is where you pick the winner and make it your production model. AcuSight uses a simple, safe promotion flow built around a champion.

​
Review your models

Open Models to see every version for the project. Select a model to view its metrics (mAP, precision, recall) and compare it against the others.
Models list

Every model version, with metrics and actions, in one place.

Open any completed model’s Metrics to see how it trained — loss, precision and recall, and mAP scores plotted across every epoch. The curves make it easy to tell whether a model converged cleanly or needs more (or cleaner) data.
Model metrics graphs

Per-model training curves — loss, precision/recall, and mAP across epochs.

​
Understand the designations

  • Champion — the project’s active production model.
  • Baseline — the previous champion, kept as a safety net to roll back to.

​
Promote a champion

  1. On the model you want to ship, select Promote to Champion.
  2. Confirm. The current champion becomes the baseline, and — if auto-deploy is on — the new champion rolls out to your devices automatically.
Emergency rollback. If a champion misbehaves in production, use Rollback to swap the champion and baseline back. The previous model redeploys to every device running the current champion.
You can also Archive models you no longer need — they’re hidden from the list but can be restored later.
Promoting a champion sets your project’s production model and, when auto-deploy is on, rolls it out to the devices already running the previous champion. To push a specific model to a specific group instead, see Deploy a model to devices.

​
Next steps

Deploy a model to devices

Ship a model to the devices you choose and watch it run.

Model lifecycle

Champion, baseline, and rollback explained.