SPECTRALook closer.
Question the finding.

Look closer.

Cortessa Spectra is a research prototype for reviewing hyperspectral measurements. It brings the scene, the spectrum, and the evidence behind a candidate finding into one workspace. In this walkthrough, we will follow a region from its initial alarm to a recorded human assessment.

Start with the scene.

We begin with a prepared research scene. This example contains plastic objects measured under controlled water conditions. The scene selector lets us move between available measurements, while the method selector chooses the detector output to inspect. An ordinary photograph alone would not provide the spectral information used here.

Choose a region.

The list on the left ranks regions for review. Selecting a region highlights its position in amber and updates the evidence panel. These are model alarms, not confirmed identifications. The surrounding scene matters: a strong response can still come from a confusing background or an unreliable calibration.

Read the spectrum.

Now look at the spectral view. Blue shows the measured spectrum; teal shows the nearest plastic training reference. The shape of these curves helps the analyst examine the proposed match. A similar curve does not, by itself, establish a polymer identity. The response score is not a probability.

Challenge the finding.

Changing the scene introduces different backgrounds and visible vegetation. Spectra keeps the selected region, its reference, and the comparator responses together. Agreement can motivate further inspection, but it is not proof. Disagreement is also useful: it points to a finding that needs a closer look rather than an automatic conclusion.

Record the judgement.

The analyst can mark a region as plastic likely, plastic unlikely, or insufficient evidence. Here, the demonstration records insufficient evidence and adds an explanatory note. That judgement belongs to the reviewer. The export control saves the review as structured data, so the assessment can be retained for follow-up.

Inspect the test.

The measured evaluation screen compares methods, detections, and false alarms. It also exposes a failed acceptance result. This matters: detecting more labelled plastic is not an improvement if the method also flags the background. These indoor research results have important limitations and do not establish reliable field performance.

Know the boundary.

Today, Spectra provides a working evidence-review workflow. Reliable material identification across independent scenes, a detection advantage over established methods, and reduced analyst workload remain to be demonstrated. The next step is qualified measured data, fair comparisons, and evaluation criteria fixed before testing.

Evidence you can inspect.

Cortessa exists to help people discover what they might otherwise miss, and understand how much they can trust what they find. Spectra carries that purpose into material analysis. Explore the research prototype at Cortessa.co.uk.