feat(perception): add diagnostic semantic SLAM replay
This commit is contained in:
DCCONSTRUCTIONS
@@ -315,6 +315,7 @@ def canonical_laboratory_adapters() -> dict[str, LaboratoryAdapter]:
|
||||
"canonical.e33-worker-shadow/v1": _run_e33,
|
||||
"canonical.e35-degradation-recovery/v1": _run_e35,
|
||||
"canonical.e46j-raw-fisheye-realtime/v1": _run_e46j,
|
||||
"experimental.e47-semantic-slam-shadow/v1": _run_e47,
|
||||
}
|
||||
|
||||
|
||||
@@ -375,6 +376,22 @@ def _run_e46j(request: LaboratoryRunRequest) -> LaboratoryAdapterResult:
|
||||
)
|
||||
|
||||
|
||||
def _run_e47(request: LaboratoryRunRequest) -> LaboratoryAdapterResult:
|
||||
from k1link.perception.semantic_slam_replay import build_semantic_slam_replay
|
||||
|
||||
result = build_semantic_slam_replay(
|
||||
repository_root=request.inputs["repository_root"],
|
||||
semantic_result_root=request.inputs["semantic_result_root"],
|
||||
threat_result_root=request.inputs["threat_result_root"],
|
||||
geometry_result_root=request.inputs["geometry_result_root"],
|
||||
output_root=request.output_root,
|
||||
)
|
||||
return LaboratoryAdapterResult(
|
||||
result_root=result.result_root,
|
||||
result_id=result.result_id,
|
||||
)
|
||||
|
||||
|
||||
def _validate_request(
|
||||
request: LaboratoryRunRequest,
|
||||
definition: LaboratoryExecutionDefinition,
|
||||
|
||||
@@ -83,6 +83,22 @@ class GeometryFrame:
|
||||
return int(self.points_map.shape[0])
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class RecordedFrameTemporalBinding:
|
||||
"""Digest-bound recorded timing evidence for one camera-indexed increment.
|
||||
|
||||
The shared session time binds the camera ordinal to the E10 pack entry. The
|
||||
LiDAR and pose deltas retain their admitted E6 meaning: nearest host-arrival
|
||||
best effort, not hardware synchronization.
|
||||
"""
|
||||
|
||||
frame_index: int
|
||||
source_time_ns: int
|
||||
source_available: bool
|
||||
lidar_camera_delta_ms: float | None
|
||||
pose_point_delta_ms: float | None
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class ReplayBodyFrameInputs:
|
||||
"""Verified inputs required to derive one replay-only virtual body frame."""
|
||||
@@ -222,13 +238,29 @@ class RecordedGeometryStore:
|
||||
or pose_reference.frame_index != envelope.sequence
|
||||
):
|
||||
raise GeometryProviderError("packet geometry references are not source-bound")
|
||||
frame_index = envelope.sequence
|
||||
frame = self.frame_for_index(envelope.sequence)
|
||||
if frame is None:
|
||||
raise GeometryProviderError("packet claims unavailable source geometry as current")
|
||||
return frame
|
||||
|
||||
def frame_for_index(self, frame_index: int) -> GeometryFrame | None:
|
||||
"""Expose one verified source increment with its pose and KB4 calibration.
|
||||
|
||||
This read-only seam is intentionally narrower than the source archive. It
|
||||
exists for deterministic replay diagnostics which must project the exact
|
||||
frame-local point index space without manufacturing a ``SourcePacket``.
|
||||
An unavailable recorded increment remains ``None``; surface validity is
|
||||
retained on the returned frame rather than silently filtering its points.
|
||||
"""
|
||||
|
||||
if not isinstance(frame_index, int) or isinstance(frame_index, bool):
|
||||
raise GeometryProviderError("replay geometry frame index is invalid")
|
||||
if not 0 <= frame_index < self.profile.frame_count:
|
||||
raise GeometryProviderError("packet geometry frame index is outside the profile")
|
||||
raise GeometryProviderError("replay geometry frame is outside the profile")
|
||||
if int(self._source["frame_indices"][frame_index]) != frame_index:
|
||||
raise GeometryProviderError("source pack frame sequence changed")
|
||||
if not bool(self._source["sample_available"][frame_index]):
|
||||
raise GeometryProviderError("packet claims unavailable source geometry as current")
|
||||
return None
|
||||
offsets = self._source["cloud_offsets"]
|
||||
start, end = int(offsets[frame_index]), int(offsets[frame_index + 1])
|
||||
return GeometryFrame(
|
||||
@@ -247,6 +279,42 @@ class RecordedGeometryStore:
|
||||
surface_valid=bool(self._surface["frame_valid"][frame_index]),
|
||||
)
|
||||
|
||||
def temporal_binding_for_index(self, frame_index: int) -> RecordedFrameTemporalBinding:
|
||||
"""Return the sealed ordinal/session binding and admitted best-effort deltas."""
|
||||
|
||||
if not isinstance(frame_index, int) or isinstance(frame_index, bool):
|
||||
raise GeometryProviderError("replay temporal frame index is invalid")
|
||||
if not 0 <= frame_index < self.profile.frame_count:
|
||||
raise GeometryProviderError("replay temporal frame is outside the profile")
|
||||
if (
|
||||
int(self._source["frame_indices"][frame_index]) != frame_index
|
||||
or int(self._source["source_frame_indices"][frame_index]) != frame_index
|
||||
):
|
||||
raise GeometryProviderError("source pack temporal sequence changed")
|
||||
session_seconds = float(self._source["session_seconds"][frame_index])
|
||||
if not math.isfinite(session_seconds) or session_seconds < 0.0:
|
||||
raise GeometryProviderError("source pack session time is invalid")
|
||||
source_available = bool(self._source["sample_available"][frame_index])
|
||||
lidar_delta = float(self._source["lidar_camera_delta_ms"][frame_index])
|
||||
pose_delta = float(self._source["pose_point_delta_ms"][frame_index])
|
||||
if source_available:
|
||||
if not math.isfinite(lidar_delta) or not math.isfinite(pose_delta):
|
||||
raise GeometryProviderError("available source temporal deltas are invalid")
|
||||
lidar_value: float | None = lidar_delta
|
||||
pose_value: float | None = pose_delta
|
||||
else:
|
||||
if not math.isnan(lidar_delta) or not math.isnan(pose_delta):
|
||||
raise GeometryProviderError("unavailable source carries temporal deltas")
|
||||
lidar_value = None
|
||||
pose_value = None
|
||||
return RecordedFrameTemporalBinding(
|
||||
frame_index=frame_index,
|
||||
source_time_ns=round(session_seconds * 1_000_000_000),
|
||||
source_available=source_available,
|
||||
lidar_camera_delta_ms=lidar_value,
|
||||
pose_point_delta_ms=pose_value,
|
||||
)
|
||||
|
||||
def current_points(self, packet: SourcePacket) -> FloatArray | None:
|
||||
"""Expose the verified frame-local point index space to temporal occupancy."""
|
||||
|
||||
@@ -393,6 +461,8 @@ class RecordedGeometryStore:
|
||||
def _validate(self) -> None:
|
||||
source_required = {
|
||||
"frame_indices",
|
||||
"source_frame_indices",
|
||||
"session_seconds",
|
||||
"sample_available",
|
||||
"cloud_offsets",
|
||||
"cloud_points_map",
|
||||
@@ -401,6 +471,8 @@ class RecordedGeometryStore:
|
||||
"intrinsic_fx_fy_cx_cy",
|
||||
"distortion_kb4",
|
||||
"t_camera_from_lidar",
|
||||
"lidar_camera_delta_ms",
|
||||
"pose_point_delta_ms",
|
||||
}
|
||||
surface_required = {"frame_valid", "point_class"}
|
||||
if not source_required.issubset(self._source):
|
||||
@@ -411,6 +483,8 @@ class RecordedGeometryStore:
|
||||
points = self.profile.point_count
|
||||
shapes = {
|
||||
"frame_indices": (frames,),
|
||||
"source_frame_indices": (frames,),
|
||||
"session_seconds": (frames,),
|
||||
"sample_available": (frames,),
|
||||
"cloud_offsets": (frames + 1,),
|
||||
"cloud_points_map": (points, 3),
|
||||
@@ -419,6 +493,8 @@ class RecordedGeometryStore:
|
||||
"intrinsic_fx_fy_cx_cy": (4,),
|
||||
"distortion_kb4": (4,),
|
||||
"t_camera_from_lidar": (4, 4),
|
||||
"lidar_camera_delta_ms": (frames,),
|
||||
"pose_point_delta_ms": (frames,),
|
||||
}
|
||||
if any(self._source[name].shape != shape for name, shape in shapes.items()):
|
||||
raise GeometryProviderError("source pack array shapes changed")
|
||||
@@ -428,6 +504,26 @@ class RecordedGeometryStore:
|
||||
raise GeometryProviderError("local surface point shape changed")
|
||||
if int(self._source["cloud_offsets"][-1]) != points:
|
||||
raise GeometryProviderError("source point offsets do not close")
|
||||
expected_indices = np.arange(frames, dtype=np.int64)
|
||||
session_seconds = np.asarray(self._source["session_seconds"], dtype=np.float64)
|
||||
if (
|
||||
not np.array_equal(self._source["frame_indices"], expected_indices)
|
||||
or not np.array_equal(self._source["source_frame_indices"], expected_indices)
|
||||
or not np.isfinite(session_seconds).all()
|
||||
or np.any(session_seconds < 0.0)
|
||||
or np.any(np.diff(session_seconds) <= 0.0)
|
||||
):
|
||||
raise GeometryProviderError("source temporal index changed")
|
||||
available = np.asarray(self._source["sample_available"], dtype=np.bool_)
|
||||
lidar_deltas = np.asarray(self._source["lidar_camera_delta_ms"], dtype=np.float64)
|
||||
pose_deltas = np.asarray(self._source["pose_point_delta_ms"], dtype=np.float64)
|
||||
if (
|
||||
not np.isfinite(lidar_deltas[available]).all()
|
||||
or not np.isfinite(pose_deltas[available]).all()
|
||||
or not np.isnan(lidar_deltas[~available]).all()
|
||||
or not np.isnan(pose_deltas[~available]).all()
|
||||
):
|
||||
raise GeometryProviderError("source temporal delta availability changed")
|
||||
if (
|
||||
int(np.count_nonzero(self._source["sample_available"]))
|
||||
!= self.profile.valid_frame_count
|
||||
@@ -1030,6 +1126,7 @@ __all__ = [
|
||||
"GeometryProviderError",
|
||||
"GeometryProviderSnapshot",
|
||||
"Ravnoves00GeometryAssociationProvider",
|
||||
"RecordedFrameTemporalBinding",
|
||||
"RecordedGeometryStore",
|
||||
"load_geometry_profile",
|
||||
]
|
||||
|
||||
@@ -0,0 +1,643 @@
|
||||
"""Model-neutral semantic diagnostics for admitted geometry observations.
|
||||
|
||||
The seam projects a provider-owned ``uint8`` semantic mask onto the existing
|
||||
frame-local point index space and aggregates those labels for already-created
|
||||
``ObstacleObservation`` values. It deliberately returns separate diagnostic
|
||||
evidence: semantic output cannot create or replace obstacle identity, metric
|
||||
occupancy, motion, threat, or safety authority.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import math
|
||||
import re
|
||||
from collections import Counter
|
||||
from dataclasses import dataclass
|
||||
from enum import IntEnum, StrEnum
|
||||
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
|
||||
from .contracts import ObstacleObservation
|
||||
from .geometry_math import ProjectedPointCloud
|
||||
|
||||
Int16Array = npt.NDArray[np.int16]
|
||||
UInt8Array = npt.NDArray[np.uint8]
|
||||
|
||||
NO_SEMANTIC_CLASS_ID = -1
|
||||
_IDENTIFIER = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._:/-]{0,159}$")
|
||||
|
||||
|
||||
class SemanticFusionError(ValueError):
|
||||
"""Semantic input or its geometry binding violates the diagnostic contract."""
|
||||
|
||||
|
||||
class SemanticClassDisposition(StrEnum):
|
||||
"""Whether one provider class is usable as a label or explicitly uncertain."""
|
||||
|
||||
LABELED = "labeled"
|
||||
AMBIGUOUS = "ambiguous"
|
||||
|
||||
|
||||
class SemanticEvidenceStatus(IntEnum):
|
||||
"""Compact source-point and observation semantic state."""
|
||||
|
||||
ABSENT = 0
|
||||
UNPROJECTED = 1
|
||||
AMBIGUOUS = 2
|
||||
LABELED = 3
|
||||
|
||||
|
||||
class SemanticEvidenceAuthority(StrEnum):
|
||||
"""Semantic output is never promoted into planner or safety authority."""
|
||||
|
||||
DIAGNOSTIC_ONLY = "diagnostic-only"
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class SemanticClassDefinition:
|
||||
"""Provider-neutral meaning assigned to one raw ``uint8`` mask value."""
|
||||
|
||||
class_id: int
|
||||
label: str
|
||||
disposition: SemanticClassDisposition = SemanticClassDisposition.LABELED
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
if (
|
||||
not isinstance(self.class_id, int)
|
||||
or isinstance(self.class_id, bool)
|
||||
or not 0 <= self.class_id <= 255
|
||||
):
|
||||
raise SemanticFusionError("semantic class id must fit uint8")
|
||||
_label(self.label, "semantic class label")
|
||||
if not isinstance(self.disposition, SemanticClassDisposition):
|
||||
raise SemanticFusionError("semantic class disposition is invalid")
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class SemanticMask:
|
||||
"""One source-bound hard semantic mask plus its complete class vocabulary."""
|
||||
|
||||
source_id: str
|
||||
frame_id: str
|
||||
provider_id: str
|
||||
model_id: str
|
||||
preprocess_id: str
|
||||
labels: UInt8Array
|
||||
classes: tuple[SemanticClassDefinition, ...]
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
for value, label in (
|
||||
(self.source_id, "semantic source id"),
|
||||
(self.frame_id, "semantic frame id"),
|
||||
(self.provider_id, "semantic provider id"),
|
||||
(self.model_id, "semantic model id"),
|
||||
(self.preprocess_id, "semantic preprocess id"),
|
||||
):
|
||||
_identifier(value, label)
|
||||
if not isinstance(self.labels, np.ndarray):
|
||||
raise SemanticFusionError("semantic mask must be a numpy array")
|
||||
if self.labels.dtype != np.uint8 or self.labels.ndim != 2:
|
||||
raise SemanticFusionError("semantic mask must have uint8 HxW shape")
|
||||
if self.labels.shape[0] < 1 or self.labels.shape[1] < 1:
|
||||
raise SemanticFusionError("semantic mask dimensions must be positive")
|
||||
if not isinstance(self.classes, tuple) or not self.classes:
|
||||
raise SemanticFusionError("semantic class vocabulary must be a nonempty tuple")
|
||||
if any(not isinstance(item, SemanticClassDefinition) for item in self.classes):
|
||||
raise SemanticFusionError("semantic class vocabulary is invalid")
|
||||
class_ids = tuple(item.class_id for item in self.classes)
|
||||
if len(set(class_ids)) != len(class_ids):
|
||||
raise SemanticFusionError("semantic class ids must be unique")
|
||||
undeclared = set(int(value) for value in np.unique(self.labels)) - set(class_ids)
|
||||
if undeclared:
|
||||
raise SemanticFusionError("semantic mask contains undeclared class ids")
|
||||
frozen = np.array(self.labels, dtype=np.uint8, order="C", copy=True)
|
||||
frozen.setflags(write=False)
|
||||
object.__setattr__(self, "labels", frozen)
|
||||
|
||||
@property
|
||||
def height(self) -> int:
|
||||
return int(self.labels.shape[0])
|
||||
|
||||
@property
|
||||
def width(self) -> int:
|
||||
return int(self.labels.shape[1])
|
||||
|
||||
def class_definition(self, class_id: int) -> SemanticClassDefinition:
|
||||
for definition in self.classes:
|
||||
if definition.class_id == class_id:
|
||||
return definition
|
||||
raise SemanticFusionError("semantic class id is not declared")
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class PointSemanticLabels:
|
||||
"""Semantic labels aligned to the complete source-point index space.
|
||||
|
||||
``class_ids`` uses ``-1`` only when the corresponding status is ``ABSENT``
|
||||
or ``UNPROJECTED``. Callers must never interpret that sentinel as a model
|
||||
class. Ambiguous provider classes retain their raw class id for review.
|
||||
"""
|
||||
|
||||
class_ids: Int16Array
|
||||
status_codes: UInt8Array
|
||||
classes: tuple[SemanticClassDefinition, ...]
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
if not isinstance(self.class_ids, np.ndarray) or self.class_ids.dtype != np.int16:
|
||||
raise SemanticFusionError("point semantic class ids must be int16")
|
||||
if self.class_ids.ndim != 1:
|
||||
raise SemanticFusionError("point semantic class ids must be one-dimensional")
|
||||
if not isinstance(self.status_codes, np.ndarray) or self.status_codes.dtype != np.uint8:
|
||||
raise SemanticFusionError("point semantic status codes must be uint8")
|
||||
if self.status_codes.shape != self.class_ids.shape:
|
||||
raise SemanticFusionError("point semantic arrays must have equal shape")
|
||||
if not isinstance(self.classes, tuple) or any(
|
||||
not isinstance(item, SemanticClassDefinition) for item in self.classes
|
||||
):
|
||||
raise SemanticFusionError("point semantic vocabulary is invalid")
|
||||
if len({item.class_id for item in self.classes}) != len(self.classes):
|
||||
raise SemanticFusionError("point semantic class ids must be unique")
|
||||
valid_statuses = {int(status) for status in SemanticEvidenceStatus}
|
||||
if set(int(value) for value in np.unique(self.status_codes)) - valid_statuses:
|
||||
raise SemanticFusionError("point semantic status code is invalid")
|
||||
unavailable = np.isin(
|
||||
self.status_codes,
|
||||
(SemanticEvidenceStatus.ABSENT, SemanticEvidenceStatus.UNPROJECTED),
|
||||
)
|
||||
if np.any(self.class_ids[unavailable] != NO_SEMANTIC_CLASS_ID):
|
||||
raise SemanticFusionError("unavailable point semantics cannot carry a class id")
|
||||
available = ~unavailable
|
||||
if np.any((self.class_ids[available] < 0) | (self.class_ids[available] > 255)):
|
||||
raise SemanticFusionError("available point semantic class id is invalid")
|
||||
definitions = {item.class_id: item for item in self.classes}
|
||||
for class_id, status_code in zip(
|
||||
self.class_ids[available].tolist(),
|
||||
self.status_codes[available].tolist(),
|
||||
strict=True,
|
||||
):
|
||||
definition = definitions.get(int(class_id))
|
||||
if definition is None:
|
||||
raise SemanticFusionError("point semantic class id is not declared")
|
||||
expected = (
|
||||
SemanticEvidenceStatus.AMBIGUOUS
|
||||
if definition.disposition is SemanticClassDisposition.AMBIGUOUS
|
||||
else SemanticEvidenceStatus.LABELED
|
||||
)
|
||||
if int(status_code) != int(expected):
|
||||
raise SemanticFusionError("point semantic status disagrees with its class")
|
||||
class_ids = np.array(self.class_ids, dtype=np.int16, order="C", copy=True)
|
||||
status_codes = np.array(self.status_codes, dtype=np.uint8, order="C", copy=True)
|
||||
class_ids.setflags(write=False)
|
||||
status_codes.setflags(write=False)
|
||||
object.__setattr__(self, "class_ids", class_ids)
|
||||
object.__setattr__(self, "status_codes", status_codes)
|
||||
|
||||
@property
|
||||
def source_point_count(self) -> int:
|
||||
return int(self.class_ids.size)
|
||||
|
||||
def status_for(self, source_point_id: int) -> SemanticEvidenceStatus:
|
||||
_point_id(source_point_id, self.source_point_count)
|
||||
return SemanticEvidenceStatus(int(self.status_codes[source_point_id]))
|
||||
|
||||
def class_id_for(self, source_point_id: int) -> int | None:
|
||||
_point_id(source_point_id, self.source_point_count)
|
||||
value = int(self.class_ids[source_point_id])
|
||||
return None if value == NO_SEMANTIC_CLASS_ID else value
|
||||
|
||||
def label_for(self, source_point_id: int) -> str | None:
|
||||
class_id = self.class_id_for(source_point_id)
|
||||
if class_id is None:
|
||||
return None
|
||||
for definition in self.classes:
|
||||
if definition.class_id == class_id:
|
||||
return definition.label
|
||||
raise AssertionError("validated point class disappeared from its vocabulary")
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class SemanticClassEvidence:
|
||||
"""Point support for one semantic class inside an existing observation."""
|
||||
|
||||
class_id: int
|
||||
label: str
|
||||
disposition: SemanticClassDisposition
|
||||
point_count: int
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
if (
|
||||
not isinstance(self.class_id, int)
|
||||
or isinstance(self.class_id, bool)
|
||||
or not 0 <= self.class_id <= 255
|
||||
):
|
||||
raise SemanticFusionError("semantic evidence class id must fit uint8")
|
||||
_label(self.label, "semantic evidence label")
|
||||
if not isinstance(self.disposition, SemanticClassDisposition):
|
||||
raise SemanticFusionError("semantic evidence disposition is invalid")
|
||||
if (
|
||||
not isinstance(self.point_count, int)
|
||||
or isinstance(self.point_count, bool)
|
||||
or self.point_count < 1
|
||||
):
|
||||
raise SemanticFusionError("semantic evidence point count must be positive")
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class ObservationSemanticEvidence:
|
||||
"""Aggregated, non-authoritative semantics for one immutable observation."""
|
||||
|
||||
observation_id: str
|
||||
occupancy_key: str
|
||||
status: SemanticEvidenceStatus
|
||||
source_point_count: int
|
||||
labeled_point_count: int
|
||||
ambiguous_point_count: int
|
||||
unprojected_point_count: int
|
||||
absent_point_count: int
|
||||
class_evidence: tuple[SemanticClassEvidence, ...]
|
||||
dominant_class_id: int | None
|
||||
dominant_label: str | None
|
||||
dominant_fraction_of_labeled: float | None
|
||||
reason_code: str
|
||||
authority: SemanticEvidenceAuthority = SemanticEvidenceAuthority.DIAGNOSTIC_ONLY
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
_identifier(self.observation_id, "semantic observation id")
|
||||
_identifier(self.occupancy_key, "semantic occupancy binding")
|
||||
_identifier(self.reason_code, "semantic evidence reason")
|
||||
if not isinstance(self.status, SemanticEvidenceStatus):
|
||||
raise SemanticFusionError("observation semantic status is invalid")
|
||||
if self.authority is not SemanticEvidenceAuthority.DIAGNOSTIC_ONLY:
|
||||
raise SemanticFusionError("semantic evidence cannot acquire product authority")
|
||||
counts = (
|
||||
self.source_point_count,
|
||||
self.labeled_point_count,
|
||||
self.ambiguous_point_count,
|
||||
self.unprojected_point_count,
|
||||
self.absent_point_count,
|
||||
)
|
||||
if any(
|
||||
not isinstance(value, int) or isinstance(value, bool) or value < 0
|
||||
for value in counts
|
||||
):
|
||||
raise SemanticFusionError("semantic evidence counts must be nonnegative integers")
|
||||
if sum(counts[1:]) != self.source_point_count:
|
||||
raise SemanticFusionError("semantic evidence accounting is incomplete")
|
||||
if not isinstance(self.class_evidence, tuple) or any(
|
||||
not isinstance(item, SemanticClassEvidence) for item in self.class_evidence
|
||||
):
|
||||
raise SemanticFusionError("semantic class evidence is invalid")
|
||||
if sum(item.point_count for item in self.class_evidence) != (
|
||||
self.labeled_point_count + self.ambiguous_point_count
|
||||
):
|
||||
raise SemanticFusionError("semantic class evidence accounting is incomplete")
|
||||
if len({item.class_id for item in self.class_evidence}) != len(self.class_evidence):
|
||||
raise SemanticFusionError("semantic class evidence ids must be unique")
|
||||
if self.status is SemanticEvidenceStatus.ABSENT:
|
||||
if self.absent_point_count != self.source_point_count:
|
||||
raise SemanticFusionError("absent semantic evidence accounting is invalid")
|
||||
elif self.status is SemanticEvidenceStatus.UNPROJECTED:
|
||||
if self.source_point_count and self.unprojected_point_count != self.source_point_count:
|
||||
raise SemanticFusionError("unprojected semantic evidence accounting is invalid")
|
||||
elif self.status is SemanticEvidenceStatus.AMBIGUOUS:
|
||||
if not self.labeled_point_count and not self.ambiguous_point_count:
|
||||
raise SemanticFusionError("ambiguous semantic evidence needs projected labels")
|
||||
elif not self.labeled_point_count:
|
||||
raise SemanticFusionError("labeled semantic evidence needs labeled points")
|
||||
dominant_values = (
|
||||
self.dominant_class_id,
|
||||
self.dominant_label,
|
||||
self.dominant_fraction_of_labeled,
|
||||
)
|
||||
if self.status is SemanticEvidenceStatus.LABELED:
|
||||
if any(value is None for value in dominant_values):
|
||||
raise SemanticFusionError("labeled semantic evidence needs a dominant class")
|
||||
if (
|
||||
not isinstance(self.dominant_fraction_of_labeled, float)
|
||||
or not math.isfinite(self.dominant_fraction_of_labeled)
|
||||
or not 0.5 < self.dominant_fraction_of_labeled <= 1.0
|
||||
):
|
||||
raise SemanticFusionError("dominant semantic fraction must be a majority")
|
||||
elif any(value is not None for value in dominant_values):
|
||||
raise SemanticFusionError("non-labeled semantic evidence cannot claim a dominant class")
|
||||
|
||||
@property
|
||||
def semantic_coverage_fraction(self) -> float:
|
||||
if not self.source_point_count:
|
||||
return 0.0
|
||||
return (self.labeled_point_count + self.ambiguous_point_count) / self.source_point_count
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class SemanticFusionResult:
|
||||
"""One frame's detached semantic diagnostics."""
|
||||
|
||||
mask_available: bool
|
||||
point_labels: PointSemanticLabels
|
||||
observation_evidence: tuple[ObservationSemanticEvidence, ...]
|
||||
authority: SemanticEvidenceAuthority = SemanticEvidenceAuthority.DIAGNOSTIC_ONLY
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
if not isinstance(self.mask_available, bool):
|
||||
raise SemanticFusionError("semantic mask availability must be boolean")
|
||||
if not isinstance(self.point_labels, PointSemanticLabels):
|
||||
raise SemanticFusionError("semantic point labels are invalid")
|
||||
if not isinstance(self.observation_evidence, tuple) or any(
|
||||
not isinstance(item, ObservationSemanticEvidence)
|
||||
for item in self.observation_evidence
|
||||
):
|
||||
raise SemanticFusionError("observation semantic evidence is invalid")
|
||||
if len({item.observation_id for item in self.observation_evidence}) != len(
|
||||
self.observation_evidence
|
||||
):
|
||||
raise SemanticFusionError("observation semantic evidence ids must be unique")
|
||||
if self.authority is not SemanticEvidenceAuthority.DIAGNOSTIC_ONLY:
|
||||
raise SemanticFusionError("semantic fusion cannot acquire product authority")
|
||||
if self.mask_available is not bool(self.point_labels.classes):
|
||||
raise SemanticFusionError("semantic mask availability and vocabulary disagree")
|
||||
|
||||
|
||||
def fuse_semantic_diagnostics(
|
||||
*,
|
||||
semantic_mask: SemanticMask | None,
|
||||
projected: ProjectedPointCloud,
|
||||
observations: tuple[ObstacleObservation, ...],
|
||||
) -> SemanticFusionResult:
|
||||
"""Attach mask diagnostics to points and observations without mutating authority."""
|
||||
|
||||
if semantic_mask is not None and not isinstance(semantic_mask, SemanticMask):
|
||||
raise SemanticFusionError("semantic mask contract is invalid")
|
||||
if not isinstance(observations, tuple) or any(
|
||||
not isinstance(item, ObstacleObservation) for item in observations
|
||||
):
|
||||
raise SemanticFusionError("geometry observations must be a tuple")
|
||||
_validate_projection(projected)
|
||||
_validate_observation_bindings(
|
||||
observations,
|
||||
semantic_mask=semantic_mask,
|
||||
source_point_count=projected.source_point_count,
|
||||
)
|
||||
point_labels = _project_point_labels(semantic_mask, projected)
|
||||
evidence = tuple(
|
||||
_aggregate_observation(observation, point_labels) for observation in observations
|
||||
)
|
||||
return SemanticFusionResult(
|
||||
mask_available=semantic_mask is not None,
|
||||
point_labels=point_labels,
|
||||
observation_evidence=evidence,
|
||||
)
|
||||
|
||||
|
||||
def _project_point_labels(
|
||||
semantic_mask: SemanticMask | None,
|
||||
projected: ProjectedPointCloud,
|
||||
) -> PointSemanticLabels:
|
||||
point_count = projected.source_point_count
|
||||
class_ids = np.full(point_count, NO_SEMANTIC_CLASS_ID, dtype=np.int16)
|
||||
if semantic_mask is None:
|
||||
return PointSemanticLabels(
|
||||
class_ids=class_ids,
|
||||
status_codes=np.full(
|
||||
point_count,
|
||||
int(SemanticEvidenceStatus.ABSENT),
|
||||
dtype=np.uint8,
|
||||
),
|
||||
classes=(),
|
||||
)
|
||||
|
||||
statuses = np.full(
|
||||
point_count,
|
||||
int(SemanticEvidenceStatus.UNPROJECTED),
|
||||
dtype=np.uint8,
|
||||
)
|
||||
if projected.projected_point_count:
|
||||
pixel_indices = np.floor(projected.pixels_xy).astype(np.int64)
|
||||
inside = (
|
||||
(pixel_indices[:, 0] >= 0)
|
||||
& (pixel_indices[:, 0] < semantic_mask.width)
|
||||
& (pixel_indices[:, 1] >= 0)
|
||||
& (pixel_indices[:, 1] < semantic_mask.height)
|
||||
)
|
||||
rows = np.flatnonzero(inside)
|
||||
if rows.size:
|
||||
source_ids = projected.source_indices[rows]
|
||||
pixels = pixel_indices[rows]
|
||||
raw_classes = semantic_mask.labels[pixels[:, 1], pixels[:, 0]]
|
||||
class_ids[source_ids] = raw_classes.astype(np.int16, copy=False)
|
||||
ambiguous_ids = np.asarray(
|
||||
[
|
||||
item.class_id
|
||||
for item in semantic_mask.classes
|
||||
if item.disposition is SemanticClassDisposition.AMBIGUOUS
|
||||
],
|
||||
dtype=np.uint8,
|
||||
)
|
||||
ambiguous = np.isin(raw_classes, ambiguous_ids)
|
||||
statuses[source_ids] = np.where(
|
||||
ambiguous,
|
||||
int(SemanticEvidenceStatus.AMBIGUOUS),
|
||||
int(SemanticEvidenceStatus.LABELED),
|
||||
).astype(np.uint8, copy=False)
|
||||
return PointSemanticLabels(
|
||||
class_ids=class_ids,
|
||||
status_codes=statuses,
|
||||
classes=semantic_mask.classes,
|
||||
)
|
||||
|
||||
|
||||
def _aggregate_observation(
|
||||
observation: ObstacleObservation,
|
||||
point_labels: PointSemanticLabels,
|
||||
) -> ObservationSemanticEvidence:
|
||||
point_ids = np.asarray(observation.source_point_ids, dtype=np.int64)
|
||||
statuses = point_labels.status_codes[point_ids]
|
||||
class_ids = point_labels.class_ids[point_ids]
|
||||
counts = Counter(int(value) for value in statuses.tolist())
|
||||
labeled_count = counts[int(SemanticEvidenceStatus.LABELED)]
|
||||
ambiguous_count = counts[int(SemanticEvidenceStatus.AMBIGUOUS)]
|
||||
unprojected_count = counts[int(SemanticEvidenceStatus.UNPROJECTED)]
|
||||
absent_count = counts[int(SemanticEvidenceStatus.ABSENT)]
|
||||
definitions = {item.class_id: item for item in point_labels.classes}
|
||||
semantic_class_ids = class_ids[
|
||||
np.isin(
|
||||
statuses,
|
||||
(SemanticEvidenceStatus.LABELED, SemanticEvidenceStatus.AMBIGUOUS),
|
||||
)
|
||||
]
|
||||
class_counts = Counter(int(value) for value in semantic_class_ids.tolist())
|
||||
class_evidence = tuple(
|
||||
SemanticClassEvidence(
|
||||
class_id=class_id,
|
||||
label=definitions[class_id].label,
|
||||
disposition=definitions[class_id].disposition,
|
||||
point_count=point_count,
|
||||
)
|
||||
for class_id, point_count in sorted(class_counts.items())
|
||||
)
|
||||
|
||||
status: SemanticEvidenceStatus
|
||||
dominant_class_id: int | None = None
|
||||
dominant_label: str | None = None
|
||||
dominant_fraction: float | None = None
|
||||
if not point_labels.classes:
|
||||
status = SemanticEvidenceStatus.ABSENT
|
||||
reason_code = "semantic-mask-unavailable"
|
||||
elif not observation.source_point_ids or unprojected_count == len(
|
||||
observation.source_point_ids
|
||||
):
|
||||
status = SemanticEvidenceStatus.UNPROJECTED
|
||||
reason_code = (
|
||||
"observation-has-no-source-points"
|
||||
if not observation.source_point_ids
|
||||
else "observation-points-unprojected"
|
||||
)
|
||||
elif not labeled_count:
|
||||
status = SemanticEvidenceStatus.AMBIGUOUS
|
||||
reason_code = "semantic-classes-ambiguous"
|
||||
else:
|
||||
labeled_ids = class_ids[statuses == int(SemanticEvidenceStatus.LABELED)]
|
||||
labeled_counts = Counter(int(value) for value in labeled_ids.tolist())
|
||||
maximum = max(labeled_counts.values())
|
||||
candidates = [
|
||||
class_id for class_id, count in labeled_counts.items() if count == maximum
|
||||
]
|
||||
semantic_point_count = labeled_count + ambiguous_count
|
||||
if len(candidates) != 1 or maximum * 2 <= semantic_point_count:
|
||||
status = SemanticEvidenceStatus.AMBIGUOUS
|
||||
reason_code = "semantic-label-majority-ambiguous"
|
||||
else:
|
||||
status = SemanticEvidenceStatus.LABELED
|
||||
dominant_class_id = candidates[0]
|
||||
dominant_label = definitions[dominant_class_id].label
|
||||
dominant_fraction = float(maximum / labeled_count)
|
||||
reason_code = "semantic-label-majority"
|
||||
return ObservationSemanticEvidence(
|
||||
observation_id=observation.observation_id,
|
||||
occupancy_key=observation.occupancy_key,
|
||||
status=status,
|
||||
source_point_count=len(observation.source_point_ids),
|
||||
labeled_point_count=labeled_count,
|
||||
ambiguous_point_count=ambiguous_count,
|
||||
unprojected_point_count=unprojected_count,
|
||||
absent_point_count=absent_count,
|
||||
class_evidence=class_evidence,
|
||||
dominant_class_id=dominant_class_id,
|
||||
dominant_label=dominant_label,
|
||||
dominant_fraction_of_labeled=dominant_fraction,
|
||||
reason_code=reason_code,
|
||||
)
|
||||
|
||||
|
||||
def _validate_projection(projected: ProjectedPointCloud) -> None:
|
||||
if not isinstance(projected, ProjectedPointCloud):
|
||||
raise SemanticFusionError("projected point cloud contract is invalid")
|
||||
if (
|
||||
not isinstance(projected.pixels_xy, np.ndarray)
|
||||
or projected.pixels_xy.dtype != np.float64
|
||||
or projected.pixels_xy.ndim != 2
|
||||
or projected.pixels_xy.shape[1:] != (2,)
|
||||
):
|
||||
raise SemanticFusionError("projected pixels must have float64 Nx2 shape")
|
||||
count = projected.projected_point_count
|
||||
if (
|
||||
not isinstance(projected.depths_m, np.ndarray)
|
||||
or projected.depths_m.dtype != np.float64
|
||||
or projected.depths_m.shape != (count,)
|
||||
):
|
||||
raise SemanticFusionError("projected depths must have float64 N shape")
|
||||
if (
|
||||
not isinstance(projected.source_indices, np.ndarray)
|
||||
or projected.source_indices.dtype != np.int64
|
||||
or projected.source_indices.shape != (count,)
|
||||
):
|
||||
raise SemanticFusionError("projected source indices must have int64 N shape")
|
||||
for value, label in (
|
||||
(projected.source_point_count, "source point count"),
|
||||
(projected.camera_front_point_count, "camera-front point count"),
|
||||
):
|
||||
if not isinstance(value, int) or isinstance(value, bool) or value < 0:
|
||||
raise SemanticFusionError(f"{label} must be a nonnegative integer")
|
||||
if not count <= projected.camera_front_point_count <= projected.source_point_count:
|
||||
raise SemanticFusionError("projected point accounting is invalid")
|
||||
if (
|
||||
not np.isfinite(projected.pixels_xy).all()
|
||||
or not np.isfinite(projected.depths_m).all()
|
||||
or np.any(projected.depths_m <= 0.0)
|
||||
):
|
||||
raise SemanticFusionError("projected point values must be finite and in front")
|
||||
if np.any(projected.source_indices < 0) or np.any(
|
||||
projected.source_indices >= projected.source_point_count
|
||||
):
|
||||
raise SemanticFusionError("projected source point id is outside the source frame")
|
||||
if np.unique(projected.source_indices).size != count:
|
||||
raise SemanticFusionError("projected source point ids must be unique")
|
||||
|
||||
|
||||
def _validate_observation_bindings(
|
||||
observations: tuple[ObstacleObservation, ...],
|
||||
*,
|
||||
semantic_mask: SemanticMask | None,
|
||||
source_point_count: int,
|
||||
) -> None:
|
||||
observation_ids: set[str] = set()
|
||||
point_owners: dict[int, str] = {}
|
||||
source_frames = {(item.source_id, item.frame_id) for item in observations}
|
||||
if len(source_frames) > 1:
|
||||
raise SemanticFusionError("geometry observations escaped their source frame")
|
||||
for observation in observations:
|
||||
if observation.observation_id in observation_ids:
|
||||
raise SemanticFusionError("geometry observation ids must be unique")
|
||||
observation_ids.add(observation.observation_id)
|
||||
if semantic_mask is not None and (
|
||||
observation.source_id != semantic_mask.source_id
|
||||
or observation.frame_id != semantic_mask.frame_id
|
||||
):
|
||||
raise SemanticFusionError("semantic mask escaped its observation source frame")
|
||||
for point_id in observation.source_point_ids:
|
||||
if point_id >= source_point_count:
|
||||
raise SemanticFusionError("observation source point id is outside the source frame")
|
||||
previous = point_owners.setdefault(point_id, observation.observation_id)
|
||||
if previous != observation.observation_id:
|
||||
raise SemanticFusionError("source point has duplicate observation ownership")
|
||||
|
||||
|
||||
def _identifier(value: str, label: str) -> None:
|
||||
if not isinstance(value, str) or _IDENTIFIER.fullmatch(value) is None:
|
||||
raise SemanticFusionError(f"{label} is invalid")
|
||||
|
||||
|
||||
def _label(value: str, label: str) -> None:
|
||||
if (
|
||||
not isinstance(value, str)
|
||||
or not value
|
||||
or value != value.strip()
|
||||
or len(value) > 120
|
||||
or any(ord(character) < 32 for character in value)
|
||||
):
|
||||
raise SemanticFusionError(f"{label} is invalid")
|
||||
|
||||
|
||||
def _point_id(value: int, source_point_count: int) -> None:
|
||||
if (
|
||||
not isinstance(value, int)
|
||||
or isinstance(value, bool)
|
||||
or not 0 <= value < source_point_count
|
||||
):
|
||||
raise SemanticFusionError("source point id is outside the point-label frame")
|
||||
|
||||
|
||||
__all__ = [
|
||||
"NO_SEMANTIC_CLASS_ID",
|
||||
"ObservationSemanticEvidence",
|
||||
"PointSemanticLabels",
|
||||
"SemanticClassDefinition",
|
||||
"SemanticClassDisposition",
|
||||
"SemanticClassEvidence",
|
||||
"SemanticEvidenceAuthority",
|
||||
"SemanticEvidenceStatus",
|
||||
"SemanticFusionError",
|
||||
"SemanticFusionResult",
|
||||
"SemanticMask",
|
||||
"fuse_semantic_diagnostics",
|
||||
]
|
||||
File diff suppressed because it is too large
Load Diff
@@ -74,6 +74,7 @@ from k1link.web.e46i_grounding_dino_full_replay_api import (
|
||||
from k1link.web.e46j_raw_fisheye_realtime_api import (
|
||||
build_e46j_raw_fisheye_realtime_router,
|
||||
)
|
||||
from k1link.web.e47_semantic_slam_api import build_e47_semantic_slam_router
|
||||
from k1link.web.environment_api import build_environment_router
|
||||
from k1link.web.l3_pointpillars_visual_api import (
|
||||
build_l3_pointpillars_visual_router,
|
||||
@@ -777,6 +778,17 @@ app.include_router(
|
||||
),
|
||||
)
|
||||
)
|
||||
app.include_router(
|
||||
build_e47_semantic_slam_router(
|
||||
root_provider=lambda: (
|
||||
REPOSITORY_ROOT
|
||||
/ ".runtime"
|
||||
/ "compute-experiments"
|
||||
/ "e47"
|
||||
/ "semantic-slam-results"
|
||||
),
|
||||
)
|
||||
)
|
||||
app.include_router(
|
||||
build_e46e_ready_stack_router(
|
||||
root_provider=lambda: (
|
||||
|
||||
@@ -0,0 +1,556 @@
|
||||
"""Read-only LAB projection of the immutable E47 semantic/SLAM shadow result."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import copy
|
||||
import hashlib
|
||||
import json
|
||||
import re
|
||||
import zipfile
|
||||
from collections.abc import Callable
|
||||
from dataclasses import dataclass
|
||||
from functools import lru_cache
|
||||
from pathlib import Path
|
||||
from typing import Final
|
||||
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
from fastapi import APIRouter, HTTPException, Query, Response
|
||||
|
||||
from k1link.perception.semantic_fusion import SemanticEvidenceStatus
|
||||
from k1link.perception.semantic_slam_replay import (
|
||||
PUBLICATION_STATUS,
|
||||
SEMANTIC_SLAM_MANIFEST_NAME,
|
||||
SEMANTIC_SLAM_MASKS_NAME,
|
||||
SEMANTIC_SLAM_OBSERVATIONS_NAME,
|
||||
SEMANTIC_SLAM_POINTS_NAME,
|
||||
SEMANTIC_SLAM_REPORT_NAME,
|
||||
SEMANTIC_SLAM_RESULT_PREFIX,
|
||||
SEMANTIC_SLAM_TAXONOMY_NAME,
|
||||
SEMANTIC_SLAM_TAXONOMY_SCHEMA,
|
||||
SemanticSlamReplayError,
|
||||
SemanticSlamReplayResult,
|
||||
read_semantic_slam_replay_result,
|
||||
)
|
||||
|
||||
E47_SEMANTIC_SLAM_CATALOG_SCHEMA: Final = "missioncore.e47-semantic-slam-catalog/v1"
|
||||
E47_SEMANTIC_SLAM_VIEW_SCHEMA: Final = "missioncore.e47-semantic-slam-view/v1"
|
||||
E47_SEMANTIC_SLAM_CHUNK_SCHEMA: Final = "missioncore.e47-semantic-slam-chunk/v1"
|
||||
E47_SEMANTIC_SLAM_FRAME_SCHEMA: Final = "missioncore.e47-semantic-slam-frame/v1"
|
||||
E47_SEMANTIC_SLAM_VIEW_STATUS: Final = "diagnostic-semantic-slam-shadow"
|
||||
E47_SEMANTIC_SLAM_MAX_CHUNK_FRAMES: Final = 24
|
||||
|
||||
_RESULT_ID = re.compile(rf"^{SEMANTIC_SLAM_RESULT_PREFIX}[a-f0-9]{{64}}$")
|
||||
_EXPECTED_ARTIFACTS: Final = (
|
||||
SEMANTIC_SLAM_MANIFEST_NAME,
|
||||
SEMANTIC_SLAM_REPORT_NAME,
|
||||
SEMANTIC_SLAM_POINTS_NAME,
|
||||
SEMANTIC_SLAM_OBSERVATIONS_NAME,
|
||||
SEMANTIC_SLAM_MASKS_NAME,
|
||||
SEMANTIC_SLAM_TAXONOMY_NAME,
|
||||
)
|
||||
_MAX_TAXONOMY_BYTES: Final = 1024 * 1024
|
||||
_MAX_MASK_BYTES: Final = 16 * 1024 * 1024
|
||||
|
||||
RootProvider = Callable[[], Path | None]
|
||||
Int64Array = npt.NDArray[np.int64]
|
||||
Int32Array = npt.NDArray[np.int32]
|
||||
UInt8Array = npt.NDArray[np.uint8]
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class _SemanticPointLedger:
|
||||
frame_offsets: Int64Array
|
||||
point_labels: UInt8Array
|
||||
point_status_codes: UInt8Array
|
||||
frame_source_point_counts: Int32Array
|
||||
frame_labeled_point_counts: Int32Array
|
||||
frame_ambiguous_point_counts: Int32Array
|
||||
frame_unprojected_point_counts: Int32Array
|
||||
frame_absent_point_counts: Int32Array
|
||||
|
||||
@property
|
||||
def frame_count(self) -> int:
|
||||
return int(self.frame_offsets.size - 1)
|
||||
|
||||
|
||||
def build_e47_semantic_slam_router(
|
||||
*,
|
||||
root_provider: RootProvider = lambda: None,
|
||||
) -> APIRouter:
|
||||
"""Expose immutable E47 evidence without granting it safety authority."""
|
||||
|
||||
router = APIRouter(
|
||||
prefix="/api/v1/laboratory/e47-semantic-slam",
|
||||
tags=["laboratory"],
|
||||
)
|
||||
|
||||
def result(result_id: str) -> SemanticSlamReplayResult:
|
||||
if _RESULT_ID.fullmatch(result_id) is None:
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден")
|
||||
root = _configured_root(root_provider)
|
||||
if root is None:
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден")
|
||||
candidate = root / result_id
|
||||
if candidate.is_symlink():
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден")
|
||||
try:
|
||||
path = candidate.resolve(strict=True)
|
||||
except OSError:
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден") from None
|
||||
if path.parent != root or not path.is_dir():
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден")
|
||||
try:
|
||||
return _read_semantic_result_cached(str(path), _result_signature(path))
|
||||
except (SemanticSlamReplayError, OSError, ValueError, zipfile.BadZipFile):
|
||||
raise HTTPException(status_code=404, detail="E47 result не найден") from None
|
||||
|
||||
def point_ledger(result_id: str) -> tuple[SemanticSlamReplayResult, _SemanticPointLedger]:
|
||||
frozen = result(result_id)
|
||||
try:
|
||||
signature = _result_signature(frozen.result_root)
|
||||
return frozen, _read_point_ledger_cached(str(frozen.result_root), signature)
|
||||
except (SemanticSlamReplayError, OSError, ValueError, zipfile.BadZipFile):
|
||||
raise HTTPException(
|
||||
status_code=503,
|
||||
detail="E47 semantic timeline не прошёл проверку",
|
||||
) from None
|
||||
|
||||
@router.get("/results")
|
||||
def list_results(limit: int = Query(default=1, ge=1, le=10)) -> dict[str, object]:
|
||||
candidates = _candidates(root_provider)
|
||||
items: list[dict[str, object]] = []
|
||||
invalid_total = 0
|
||||
for candidate in candidates:
|
||||
if len(items) >= limit:
|
||||
break
|
||||
try:
|
||||
items.append(_project_result(result(candidate.name)))
|
||||
except (HTTPException, OSError, ValueError, json.JSONDecodeError):
|
||||
invalid_total += 1
|
||||
return {
|
||||
"schema_version": E47_SEMANTIC_SLAM_CATALOG_SCHEMA,
|
||||
"configured": _configured_root(root_provider) is not None,
|
||||
"items": items,
|
||||
"candidate_total": len(candidates),
|
||||
"invalid_total": invalid_total,
|
||||
"access": "read-only-diagnostic-shadow",
|
||||
}
|
||||
|
||||
@router.get("/results/{result_id}/timeline/chunk")
|
||||
def get_timeline_chunk(
|
||||
result_id: str,
|
||||
start: int = Query(default=0, ge=0),
|
||||
count: int = Query(
|
||||
default=12,
|
||||
ge=1,
|
||||
le=E47_SEMANTIC_SLAM_MAX_CHUNK_FRAMES,
|
||||
),
|
||||
) -> dict[str, object]:
|
||||
frozen, ledger = point_ledger(result_id)
|
||||
if (
|
||||
not isinstance(start, int)
|
||||
or isinstance(start, bool)
|
||||
or not isinstance(count, int)
|
||||
or isinstance(count, bool)
|
||||
or start < 0
|
||||
or not 1 <= count <= E47_SEMANTIC_SLAM_MAX_CHUNK_FRAMES
|
||||
):
|
||||
raise HTTPException(status_code=422, detail="Некорректный E47 timeline chunk")
|
||||
if start >= ledger.frame_count:
|
||||
raise HTTPException(status_code=404, detail="E47 timeline chunk не найден")
|
||||
stop = min(start + count, ledger.frame_count)
|
||||
try:
|
||||
frames = [_project_frame(ledger, sequence) for sequence in range(start, stop)]
|
||||
except ValueError:
|
||||
raise HTTPException(
|
||||
status_code=503,
|
||||
detail="E47 semantic timeline не прошёл проверку",
|
||||
) from None
|
||||
return {
|
||||
"schema_version": E47_SEMANTIC_SLAM_CHUNK_SCHEMA,
|
||||
"result_id": frozen.result_id,
|
||||
"start_sequence": start,
|
||||
"frame_count": len(frames),
|
||||
"next_sequence": stop if stop < ledger.frame_count else None,
|
||||
"frames": frames,
|
||||
"access": "read-only-diagnostic-shadow",
|
||||
}
|
||||
|
||||
@router.get("/results/{result_id}/masks/{sequence}")
|
||||
def get_mask(result_id: str, sequence: int) -> Response:
|
||||
frozen = result(result_id)
|
||||
frame_total = _frame_total(frozen)
|
||||
if (
|
||||
not isinstance(sequence, int)
|
||||
or isinstance(sequence, bool)
|
||||
or not 0 <= sequence < frame_total
|
||||
):
|
||||
raise HTTPException(status_code=404, detail="E47 semantic mask не найдена")
|
||||
try:
|
||||
signature = _result_signature(frozen.result_root)
|
||||
frozen = _read_semantic_result_cached(str(frozen.result_root), signature)
|
||||
payload = _read_mask(frozen, sequence)
|
||||
if _result_signature(frozen.result_root) != signature:
|
||||
raise ValueError("E47 result changed during mask read")
|
||||
except (
|
||||
SemanticSlamReplayError,
|
||||
OSError,
|
||||
KeyError,
|
||||
ValueError,
|
||||
RuntimeError,
|
||||
zipfile.BadZipFile,
|
||||
):
|
||||
raise HTTPException(
|
||||
status_code=503,
|
||||
detail="E47 semantic mask не прошла проверку",
|
||||
) from None
|
||||
digest = hashlib.sha256(payload).hexdigest()
|
||||
return Response(
|
||||
content=payload,
|
||||
media_type="image/png",
|
||||
headers={
|
||||
"Cache-Control": "private, max-age=31536000, immutable",
|
||||
"ETag": f'"{digest}"',
|
||||
"X-Content-Type-Options": "nosniff",
|
||||
},
|
||||
)
|
||||
|
||||
return router
|
||||
|
||||
|
||||
@lru_cache(maxsize=4)
|
||||
def _read_semantic_result_cached(
|
||||
root_value: str,
|
||||
signature: tuple[int, ...],
|
||||
) -> SemanticSlamReplayResult:
|
||||
del signature
|
||||
return read_semantic_slam_replay_result(Path(root_value))
|
||||
|
||||
|
||||
@lru_cache(maxsize=2)
|
||||
def _read_point_ledger_cached(
|
||||
root_value: str,
|
||||
signature: tuple[int, ...],
|
||||
) -> _SemanticPointLedger:
|
||||
frozen = _read_semantic_result_cached(root_value, signature)
|
||||
path = frozen.result_root / SEMANTIC_SLAM_POINTS_NAME
|
||||
required = {
|
||||
"frame_offsets",
|
||||
"point_labels",
|
||||
"point_status_codes",
|
||||
"frame_source_point_counts",
|
||||
"frame_labeled_point_counts",
|
||||
"frame_ambiguous_point_counts",
|
||||
"frame_unprojected_point_counts",
|
||||
"frame_absent_point_counts",
|
||||
}
|
||||
with np.load(path, allow_pickle=False) as archive:
|
||||
if not required.issubset(archive.files):
|
||||
raise ValueError("E47 semantic point arrays are incomplete")
|
||||
ledger = _SemanticPointLedger(
|
||||
frame_offsets=_frozen_int64(archive["frame_offsets"]),
|
||||
point_labels=_frozen_uint8(archive["point_labels"]),
|
||||
point_status_codes=_frozen_uint8(archive["point_status_codes"]),
|
||||
frame_source_point_counts=_frozen_int32(archive["frame_source_point_counts"]),
|
||||
frame_labeled_point_counts=_frozen_int32(archive["frame_labeled_point_counts"]),
|
||||
frame_ambiguous_point_counts=_frozen_int32(archive["frame_ambiguous_point_counts"]),
|
||||
frame_unprojected_point_counts=_frozen_int32(archive["frame_unprojected_point_counts"]),
|
||||
frame_absent_point_counts=_frozen_int32(archive["frame_absent_point_counts"]),
|
||||
)
|
||||
_validate_point_ledger(ledger, _frame_total(frozen))
|
||||
_validate_class_status_bindings(ledger, _read_taxonomy(frozen))
|
||||
return ledger
|
||||
|
||||
|
||||
def _project_result(result: SemanticSlamReplayResult) -> dict[str, object]:
|
||||
if result.status != PUBLICATION_STATUS:
|
||||
raise ValueError("E47 publication status changed")
|
||||
identity = _object(result.manifest.get("identity"), "E47 identity")
|
||||
provider = _object(identity.get("semantic_provider"), "E47 provider")
|
||||
temporal_binding = _object(
|
||||
identity.get("temporal_binding"),
|
||||
"E47 temporal binding",
|
||||
)
|
||||
authority = _object(identity.get("authority"), "E47 authority")
|
||||
if (
|
||||
authority.get("ground_truth") is not False
|
||||
or authority.get("semantic_authority") != "diagnostic-only"
|
||||
or authority.get("navigation_or_safety_accepted") is not False
|
||||
or authority.get("actuation_allowed") is not False
|
||||
):
|
||||
raise ValueError("E47 authority changed")
|
||||
taxonomy = _read_taxonomy(result)
|
||||
return {
|
||||
"schema_version": E47_SEMANTIC_SLAM_VIEW_SCHEMA,
|
||||
"result_id": result.result_id,
|
||||
"created_at_utc": result.manifest["created_at_utc"],
|
||||
"status": E47_SEMANTIC_SLAM_VIEW_STATUS,
|
||||
"profile_id": identity["profile_id"],
|
||||
"base_m4_result_id": identity["base_m4_result_id"],
|
||||
"semantic_result_id": identity["semantic_result_id"],
|
||||
"geometry_result_id": identity["geometry_result_id"],
|
||||
"source_pack_id": identity["source_pack_id"],
|
||||
"calibration_content_sha256": identity["calibration_content_sha256"],
|
||||
"provider": {
|
||||
"provider_id": provider["provider_id"],
|
||||
"model_id": provider["model_id"],
|
||||
"model_revision": provider["model_revision"],
|
||||
"model_weights_sha256": provider["model_weights_sha256"],
|
||||
"preprocess_id": provider["preprocess_id"],
|
||||
},
|
||||
"temporal_binding": copy.deepcopy(temporal_binding),
|
||||
"taxonomy": taxonomy,
|
||||
"metrics": copy.deepcopy(result.metrics),
|
||||
"acceptance": {
|
||||
"artifact_contract_passed": True,
|
||||
"frame_accounting_passed": True,
|
||||
"point_accounting_passed": True,
|
||||
"observation_binding_passed": True,
|
||||
"temporal_binding_passed": True,
|
||||
"independent_semantic_truth_passed": False,
|
||||
"provider_promoted": False,
|
||||
},
|
||||
"limitations": copy.deepcopy(result.report["limitations"]),
|
||||
"ground_truth": False,
|
||||
"semantic_authority": "diagnostic-only",
|
||||
"navigation_or_safety_accepted": False,
|
||||
"actuation_allowed": False,
|
||||
"access": "read-only-diagnostic-shadow",
|
||||
}
|
||||
|
||||
|
||||
def _project_frame(ledger: _SemanticPointLedger, sequence: int) -> dict[str, object]:
|
||||
offset = int(ledger.frame_offsets[sequence])
|
||||
stop = int(ledger.frame_offsets[sequence + 1])
|
||||
labels = ledger.point_labels[offset:stop].astype(np.int16)
|
||||
statuses = ledger.point_status_codes[offset:stop]
|
||||
unavailable = np.isin(
|
||||
statuses,
|
||||
(
|
||||
int(SemanticEvidenceStatus.ABSENT),
|
||||
int(SemanticEvidenceStatus.UNPROJECTED),
|
||||
),
|
||||
)
|
||||
labels[unavailable] = -1
|
||||
counts = {
|
||||
"labeled": int(ledger.frame_labeled_point_counts[sequence]),
|
||||
"ambiguous": int(ledger.frame_ambiguous_point_counts[sequence]),
|
||||
"unprojected": int(ledger.frame_unprojected_point_counts[sequence]),
|
||||
"absent": int(ledger.frame_absent_point_counts[sequence]),
|
||||
}
|
||||
actual_counts = {
|
||||
"labeled": int(np.count_nonzero(statuses == int(SemanticEvidenceStatus.LABELED))),
|
||||
"ambiguous": int(np.count_nonzero(statuses == int(SemanticEvidenceStatus.AMBIGUOUS))),
|
||||
"unprojected": int(np.count_nonzero(statuses == int(SemanticEvidenceStatus.UNPROJECTED))),
|
||||
"absent": int(np.count_nonzero(statuses == int(SemanticEvidenceStatus.ABSENT))),
|
||||
}
|
||||
if counts != actual_counts or sum(counts.values()) != stop - offset:
|
||||
raise ValueError("E47 frame point accounting changed")
|
||||
return {
|
||||
"schema_version": E47_SEMANTIC_SLAM_FRAME_SCHEMA,
|
||||
"sequence": sequence,
|
||||
"source_point_count": int(ledger.frame_source_point_counts[sequence]),
|
||||
"class_ids": labels.tolist(),
|
||||
"status_codes": statuses.tolist(),
|
||||
"counts": counts,
|
||||
}
|
||||
|
||||
|
||||
def _read_taxonomy(result: SemanticSlamReplayResult) -> list[dict[str, object]]:
|
||||
path = result.result_root / SEMANTIC_SLAM_TAXONOMY_NAME
|
||||
if not path.is_file() or path.is_symlink() or path.stat().st_size > _MAX_TAXONOMY_BYTES:
|
||||
raise ValueError("E47 taxonomy is invalid")
|
||||
payload = path.read_bytes()
|
||||
identity = _object(result.manifest.get("identity"), "E47 identity")
|
||||
if hashlib.sha256(payload).hexdigest() != identity.get("taxonomy_sha256"):
|
||||
raise ValueError("E47 taxonomy identity changed")
|
||||
document = json.loads(payload)
|
||||
if not isinstance(document, dict) or set(document) != {"schema_version", "classes"}:
|
||||
raise ValueError("E47 taxonomy contract changed")
|
||||
if document.get("schema_version") != SEMANTIC_SLAM_TAXONOMY_SCHEMA:
|
||||
raise ValueError("E47 taxonomy schema changed")
|
||||
classes = document.get("classes")
|
||||
if not isinstance(classes, list) or not classes:
|
||||
raise ValueError("E47 taxonomy classes are invalid")
|
||||
for item in classes:
|
||||
if not isinstance(item, dict) or set(item) != {
|
||||
"class_id",
|
||||
"label",
|
||||
"disposition",
|
||||
"color_rgb",
|
||||
}:
|
||||
raise ValueError("E47 taxonomy class changed")
|
||||
return copy.deepcopy(classes)
|
||||
|
||||
|
||||
def _read_mask(result: SemanticSlamReplayResult, sequence: int) -> bytes:
|
||||
archive_path = result.result_root / SEMANTIC_SLAM_MASKS_NAME
|
||||
if not archive_path.is_file() or archive_path.is_symlink():
|
||||
raise ValueError("E47 semantic mask archive is invalid")
|
||||
member_name = f"semantic-masks/frame-{sequence + 1:06d}.png"
|
||||
with zipfile.ZipFile(archive_path, mode="r") as archive:
|
||||
info = archive.getinfo(member_name)
|
||||
if info.is_dir() or not 0 < info.file_size <= _MAX_MASK_BYTES:
|
||||
raise ValueError("E47 semantic mask member is invalid")
|
||||
payload = archive.read(info)
|
||||
if len(payload) != info.file_size or not payload.startswith(b"\x89PNG\r\n\x1a\n"):
|
||||
raise ValueError("E47 semantic mask payload is invalid")
|
||||
return payload
|
||||
|
||||
|
||||
def _validate_point_ledger(ledger: _SemanticPointLedger, frame_total: int) -> None:
|
||||
arrays = (
|
||||
ledger.frame_source_point_counts,
|
||||
ledger.frame_labeled_point_counts,
|
||||
ledger.frame_ambiguous_point_counts,
|
||||
ledger.frame_unprojected_point_counts,
|
||||
ledger.frame_absent_point_counts,
|
||||
)
|
||||
if (
|
||||
ledger.frame_offsets.ndim != 1
|
||||
or ledger.frame_offsets.shape != (frame_total + 1,)
|
||||
or int(ledger.frame_offsets[0]) != 0
|
||||
or np.any(np.diff(ledger.frame_offsets) < 0)
|
||||
or ledger.point_labels.ndim != 1
|
||||
or ledger.point_status_codes.shape != ledger.point_labels.shape
|
||||
or int(ledger.frame_offsets[-1]) != ledger.point_labels.size
|
||||
or any(value.ndim != 1 or value.shape != (frame_total,) for value in arrays)
|
||||
or np.any(np.asarray(arrays) < 0)
|
||||
or not np.array_equal(
|
||||
np.diff(ledger.frame_offsets),
|
||||
ledger.frame_source_point_counts,
|
||||
)
|
||||
):
|
||||
raise ValueError("E47 semantic point ledger changed")
|
||||
valid_statuses = {int(status) for status in SemanticEvidenceStatus}
|
||||
if set(int(value) for value in np.unique(ledger.point_status_codes)) - valid_statuses:
|
||||
raise ValueError("E47 semantic status changed")
|
||||
expected_total = (
|
||||
ledger.frame_labeled_point_counts
|
||||
+ ledger.frame_ambiguous_point_counts
|
||||
+ ledger.frame_unprojected_point_counts
|
||||
+ ledger.frame_absent_point_counts
|
||||
)
|
||||
if not np.array_equal(expected_total, ledger.frame_source_point_counts):
|
||||
raise ValueError("E47 semantic frame accounting changed")
|
||||
|
||||
|
||||
def _validate_class_status_bindings(
|
||||
ledger: _SemanticPointLedger,
|
||||
taxonomy: list[dict[str, object]],
|
||||
) -> None:
|
||||
dispositions: dict[int, str] = {}
|
||||
for item in taxonomy:
|
||||
class_id = item.get("class_id")
|
||||
disposition = item.get("disposition")
|
||||
if (
|
||||
not isinstance(class_id, int)
|
||||
or isinstance(class_id, bool)
|
||||
or not 0 <= class_id <= 255
|
||||
or disposition not in {"labeled", "ambiguous"}
|
||||
or class_id in dispositions
|
||||
):
|
||||
raise ValueError("E47 semantic taxonomy binding changed")
|
||||
dispositions[class_id] = str(disposition)
|
||||
unavailable = np.isin(
|
||||
ledger.point_status_codes,
|
||||
(
|
||||
int(SemanticEvidenceStatus.ABSENT),
|
||||
int(SemanticEvidenceStatus.UNPROJECTED),
|
||||
),
|
||||
)
|
||||
if np.any(ledger.point_labels[unavailable] != 0):
|
||||
raise ValueError("E47 unavailable semantic point carried a class")
|
||||
for status, disposition in (
|
||||
(SemanticEvidenceStatus.AMBIGUOUS, "ambiguous"),
|
||||
(SemanticEvidenceStatus.LABELED, "labeled"),
|
||||
):
|
||||
class_ids = np.unique(ledger.point_labels[ledger.point_status_codes == int(status)])
|
||||
if any(dispositions.get(int(class_id)) != disposition for class_id in class_ids):
|
||||
raise ValueError("E47 semantic point status disagrees with taxonomy")
|
||||
|
||||
|
||||
def _frame_total(result: SemanticSlamReplayResult) -> int:
|
||||
frames = _object(result.metrics.get("frames"), "E47 frame metrics")
|
||||
value = frames.get("total")
|
||||
if not isinstance(value, int) or isinstance(value, bool) or value < 1:
|
||||
raise ValueError("E47 frame count changed")
|
||||
return value
|
||||
|
||||
|
||||
def _frozen_int64(value: npt.ArrayLike) -> Int64Array:
|
||||
array = np.array(value, dtype=np.int64, order="C", copy=True)
|
||||
array.setflags(write=False)
|
||||
return array
|
||||
|
||||
|
||||
def _frozen_int32(value: npt.ArrayLike) -> Int32Array:
|
||||
array = np.array(value, dtype=np.int32, order="C", copy=True)
|
||||
array.setflags(write=False)
|
||||
return array
|
||||
|
||||
|
||||
def _frozen_uint8(value: npt.ArrayLike) -> UInt8Array:
|
||||
array = np.array(value, dtype=np.uint8, order="C", copy=True)
|
||||
array.setflags(write=False)
|
||||
return array
|
||||
|
||||
|
||||
def _configured_root(provider: RootProvider) -> Path | None:
|
||||
value = provider()
|
||||
if value is None:
|
||||
return None
|
||||
candidate = value.expanduser().absolute()
|
||||
if candidate.is_symlink():
|
||||
return None
|
||||
try:
|
||||
root = candidate.resolve(strict=True)
|
||||
except OSError:
|
||||
return None
|
||||
return root if root.is_dir() else None
|
||||
|
||||
|
||||
def _result_signature(root: Path) -> tuple[int, ...]:
|
||||
signature: list[int] = []
|
||||
for name in _EXPECTED_ARTIFACTS:
|
||||
path = root / name
|
||||
if not path.is_file() or path.is_symlink():
|
||||
raise ValueError("E47 result artifact is invalid")
|
||||
stat = path.stat()
|
||||
signature.extend((stat.st_ino, stat.st_size, stat.st_mtime_ns, stat.st_ctime_ns))
|
||||
return tuple(signature)
|
||||
|
||||
|
||||
def _candidates(provider: RootProvider) -> list[Path]:
|
||||
root = _configured_root(provider)
|
||||
if root is None:
|
||||
return []
|
||||
try:
|
||||
return sorted(
|
||||
(
|
||||
item
|
||||
for item in root.iterdir()
|
||||
if item.is_dir() and not item.is_symlink() and _RESULT_ID.fullmatch(item.name)
|
||||
),
|
||||
key=lambda item: item.stat().st_mtime_ns,
|
||||
reverse=True,
|
||||
)
|
||||
except OSError:
|
||||
return []
|
||||
|
||||
|
||||
def _object(value: object, label: str) -> dict[str, object]:
|
||||
if not isinstance(value, dict):
|
||||
raise ValueError(f"{label} is invalid")
|
||||
return value
|
||||
|
||||
|
||||
__all__ = [
|
||||
"E47_SEMANTIC_SLAM_CATALOG_SCHEMA",
|
||||
"E47_SEMANTIC_SLAM_CHUNK_SCHEMA",
|
||||
"E47_SEMANTIC_SLAM_FRAME_SCHEMA",
|
||||
"E47_SEMANTIC_SLAM_MAX_CHUNK_FRAMES",
|
||||
"E47_SEMANTIC_SLAM_VIEW_SCHEMA",
|
||||
"build_e47_semantic_slam_router",
|
||||
]
|
||||
Reference in New Issue
Block a user