Reviewing a detection is two separate decisions: was the AI right, and does this belong in my inspection record. They are not the same question, and Twinsity keeps them apart.
| Action | Key | Means |
| Accept | A | The AI found something real |
| Dismiss | D | The AI found nothing real |
The keyboard shortcuts are much faster than clicking when you are working through a class.
Clicking the verdict you already set clears it, returning the detection to unreviewed. Use that when you change your mind or mis-key.
A verdict records your judgement of the AI. It does not create anything.
Add observation creates a real observation from the detection, carrying over its image and location. Pressing O does the same thing.
Choose the observation class while converting. The AI's class is a starting point, not the final answer: what it calls corrosion may be staining, and you classify it as you would any observation of your own.
Once converted, Add observation is replaced by Open observation, which takes you to it in Inspect. A detection converts once.
Delete observation removes the observation the conversion created, with a confirmation. The detection itself stays, so you can convert again.
You can accept without converting, and convert without accepting. In practice:
Accepting everything real, and converting only what belongs in the report, keeps your record clean while still telling you how the AI performed.
| What you see | What it means |
| Accept and Dismiss are greyed | You need Editor access to this project. |
| Add observation is greyed | Either you lack Editor access, or this detection is already converted. Look for Open observation. |
| No class picker while converting | The project's profile has no classes; the observation gets the default. |
| No Delete observation option | You need Editor access to undo a conversion. |
| Accept, Dismiss and Add observation are greyed in a project labelled Demo | Verdicts and conversions are not available in demo projects. |