Is this page helpful?

Safety AI Monitor (SAIM)#

Safety AI Monitor continuously assesses the health of the camera sensor inputs feeding HOISA. It independently observes the same RTSP streams consumed by SIPP and reports trust verdicts to SEI so that fusion and downstream decisions can factor sensor integrity into the safety response.

Operating modes#

SAIM Operating Modes#
Mode	Description
LEARN	Observes all configured streams for a configurable duration and records a per-sensor baseline of frame-quality metrics. No events are reported to SEI in this mode.
ACTIVE	Loads the learned baselines, registers with the SEI daemon, and continuously scores every decoded frame against its baseline. Publishes trust reports to SEI when stream health changes.

SENSOR_INVALID reporting conditions#

In ACTIVE mode, SAIM publishes SENSOR_INVALID trust reports for a given sensor in any of the following cases:

Out-of-Distribution (OOD) Detection — Identifies frames that deviate significantly from normal operating conditions.
Camera Blockage Detection — Detects physical obstruction, fogging, or complete lens coverage.
RTSP transport loss — The RTSP/RTP session cannot be established or re-established after the configured number of connection retries.
Stream corruption — H.264 FU-A fragments are dropped beyond the configured threshold, indicating packet loss or malformed streams.

When the stream recovers and its frame quality returns inside the baseline envelope, SAIM reports SENSOR_VALID to restore trust.

Note

Cold-start behavior. On every startup, SAIM emits one SENSOR_INVALID event per configured pipeline before any real frames have been processed. Once the streams are established and the first frames clear the baseline checks, SAIM emits a matching SENSOR_VALID event per pipeline to recover trust. This invalid-then-valid transient is expected and does not indicate a sensor fault.