Monty.Frameworks.Models.Mixins.NoResetEvidence.HypothesesUpdaterGraphTelemetry (ExTbpMonty.Core v0.1.0)

One graph's hypotheses, as the updater left them: what a chart draws.

Mirrors frameworks/models/mixins/no_reset_evidence.py::HypothesesUpdaterGraphTelemetry, the dataclass Monty's pyview dashboard reads. Four arrays and the updater's own telemetry, one entry per hypothesis of one graph:

evidence — what each hypothesis is worth;
locations — where each one puts the object, in metres;
rotations — how each one is turned, as Euler angles in degrees in Monty's xyz order;
pose_errors — how far each hypothesis' rotation is from the target's, in radians, which is the number a recognition's accuracy is measured by;
hypotheses_updater — whatever the updater said about this graph, passed through rather than interpreted.

Monty keeps these as numpy arrays; here they are lists, and Monty.Frameworks.Models.Buffer.BufferEncoder writes a struct of this shape as a JSON object the way Monty's own encoder writes a dataclass.

Summary

Types

t()

Functions

empty(graph_telemetry)

The telemetry of a graph that has no hypotheses.

Types

t()

@type t() ::
  %Monty.Frameworks.Models.Mixins.NoResetEvidence.HypothesesUpdaterGraphTelemetry{
    evidence: [float()],
    hypotheses_updater: map(),
    locations: [[float()]],
    pose_errors: [float()],
    rotations: [[float()]]
  }

Functions

empty(graph_telemetry)

@spec empty(map()) :: t()

The telemetry of a graph that has no hypotheses.

Monty builds the same record with four empty arrays — a graph with nothing believed still has a row in a log, and leaving it out would make "no hypotheses" look like "not reported".