DETECTION TYPES REFERENCE
What SPECTER flags, and what each label means
SPECTER classifies anomalies with a local neural network and a classic computer-vision pipeline. Everything runs on device. No frames leave the machine and there is no cloud step. Each detection is scored and stored with its confidence and a sigma class so you can sort the strong hits from the weak ones later.
THE TYPES
| Type | What triggers it |
|---|---|
| Spectral Anomaly | A general anomaly in the frame that does not fit a more specific class. |
| Orb | A small, bright, roughly circular feature moving through the scene. |
| Cold Spot | A localized low-temperature region in the thermal feed. |
| Partial Apparition | A partial human-like form that does not resolve into a full figure. |
| Vapor Anomaly | A diffuse, cloud-like mass with soft edges. |
| Chromatic Signature | An unusual, concentrated color signature against the background. |
| Energy Vortex | A rotational or swirling pattern of motion or field activity. |
| Disembodied Face | Face-like features detected with no supporting body. |
| Facial Anomaly | A face-like pattern that reads as irregular or out of place. |
| Humanoid Figure | A full upright human-like form resolved by the skeletal model. |
Every detection carries a confidence score and a sigma class in the gallery. Use them to rank your evidence and filter out the low-confidence noise.