JensandClaude Opus 5
ff4a15aa74
The review vocabulary already separates a model error from a reference gap, because the product's position is that official footprints are not automatically perfect ground truth. Those verdicts were only counted. An operator who inspected forty false positives and established that twelve are buildings the reference simply lacks still saw a precision counting all forty against the model — a number they had personally disproved, on the panel where they disproved it. Applying the verdicts gives an adjudicated score reported next to the raw one, so nothing is quietly improved. Not being able to judge is not evidence in the model's favour, so uncertain and obscured verdicts keep counting, as does a decision from a later release that this runtime does not recognise. Because part of the evidence is usually still unreviewed, the honest form is an interval rather than a single corrected number: pessimistic assumes every unreviewed finding is a model error, optimistic assumes none is, and the headline equals the pessimistic reading so a partly reviewed check never presents as a settled one. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
68 lines
2.0 KiB
Python
68 lines
2.0 KiB
Python
from __future__ import annotations
|
|
|
|
from datetime import datetime
|
|
from typing import Literal
|
|
from uuid import UUID
|
|
|
|
from pydantic import BaseModel, Field
|
|
|
|
|
|
DetectionEvidenceRole = Literal["false_positive", "false_negative"]
|
|
DetectionReviewDecision = Literal[
|
|
"confirmed_model_false_positive",
|
|
"confirmed_model_false_negative",
|
|
"reference_gap_or_change",
|
|
"qa_alignment_mismatch",
|
|
"imagery_obscured_or_uncertain",
|
|
"uncertain",
|
|
"unreviewed",
|
|
]
|
|
|
|
|
|
class DetectionReviewUpsert(BaseModel):
|
|
evidence_role: DetectionEvidenceRole
|
|
evidence_feature_id: str = Field(min_length=1, max_length=255)
|
|
decision: DetectionReviewDecision
|
|
notes: str | None = Field(default=None, max_length=2000)
|
|
reviewed_by: str = Field(default="operator", min_length=1, max_length=120)
|
|
|
|
|
|
class DetectionReviewRead(BaseModel):
|
|
id: UUID | None = None
|
|
project_id: UUID
|
|
quality_check_id: UUID
|
|
analysis_run_id: UUID | None = None
|
|
evidence_role: DetectionEvidenceRole
|
|
evidence_feature_id: str
|
|
detection_id: UUID | None = None
|
|
reference_feature_id: UUID | None = None
|
|
decision: DetectionReviewDecision = "unreviewed"
|
|
notes: str | None = None
|
|
reviewed_by: str | None = None
|
|
confidence: float | None = None
|
|
class_name: str | None = None
|
|
source_tile_path: str | None = None
|
|
created_at: datetime | None = None
|
|
updated_at: datetime | None = None
|
|
|
|
|
|
class DetectionReviewSummary(BaseModel):
|
|
total: int
|
|
reviewed: int
|
|
remaining: int
|
|
false_positive_total: int
|
|
false_negative_total: int
|
|
decision_counts: dict[str, int]
|
|
# The score with the operator's verdicts applied, next to the raw one. A
|
|
# finding adjudicated as a reference gap is not the model's error, and an
|
|
# interval covers what the unreviewed remainder could still turn out to be.
|
|
reviewed_metrics: dict | None = None
|
|
|
|
|
|
class DetectionReviewList(BaseModel):
|
|
items: list[DetectionReviewRead]
|
|
total: int
|
|
limit: int
|
|
offset: int
|
|
summary: DetectionReviewSummary
|