segrails.result¶

Segrails result dataclasses.

Defines the structured output returned by the Segrails inference pipeline.

Classes¶

`SegrailsError`	Error descriptor returned when a pipeline step fails.
`ModelInfo`	Metadata about the model used during inference.
`SegrailsPrediction`	Detection results produced by the postprocessing step.
`SegrailsResult`	Top-level result returned by the Segrails pipeline.

Module Contents¶

class segrails.result.SegrailsError¶

Error descriptor returned when a pipeline step fails.

code¶: HTTP-style error code (e.g. 500 for an internal error).

message¶: Human-readable description of the error.

code: int¶

message: str¶

class segrails.result.ModelInfo¶

Metadata about the model used during inference.

name¶: Registered model name.

version¶: Registered model version.

model_path¶: Local path to the model checkpoint file.

name: str¶

version: str¶

model_path: str¶

class segrails.result.SegrailsPrediction¶

Detection results produced by the postprocessing step.

inference_time¶: Duration of the UNet inference in seconds.

n_tracks¶: Total number of track segments detected.

valid_tracks¶: Number of tracks classified as valid rail tracks.

crossings¶: Number of segments classified as rail crossings.

total_low_detections¶: Total count of detections with IoU below threshold.

total_missed_detections¶: Total count of track segments with no detection.

tracks¶: List of track dictionaries. Each entry contains id, classification, and a list of point dicts with coordinates, direction, IoU score and status.

inference_time: float = 0.0¶

n_tracks: int = 0¶

valid_tracks: int = 0¶

crossings: int = 0¶

total_low_detections: int = 0¶

total_missed_detections: int = 0¶

tracks: List[Dict[str, Any]] = []¶

class segrails.result.SegrailsResult¶

Top-level result returned by the Segrails pipeline.

status¶: Pipeline status. Either "success" or "failure".

timestamp¶: ISO 8601 UTC timestamp of when the pipeline ran.

model¶: Metadata about the model used. None if loading failed.

input¶: Optional dict describing the pipeline inputs (paths, center, …).

predictions¶: Detection results. None if inference or postprocessing failed.

error¶: Error descriptor. None when status is "success".

Example

>>> result = SegrailsResult(status="success", timestamp="2026-01-01T00:00:00Z")
>>> result.dump()
{'status': 'success', 'timestamp': '2026-01-01T00:00:00Z', ...}

status: Literal['success', 'failure']¶

timestamp: str¶

model: ModelInfo | None = None¶

input: Dict[str, Any] | None = None¶

predictions: SegrailsPrediction | None = None¶

error: SegrailsError | None = None¶

dump() → Dict[str, Any]¶

Serialize the result to a plain dictionary.

Returns:: Fully serialized result, suitable for JSON export.
Return type:: Dict[str, Any]