Service

app.application.services.detection.service

View source code

Detection service orchestration for model-based object detection across sessions.

See Also

• DetectionEngineManager — manages per-session inference engines
• DetectionWorkerFactoryAdapter — creates background detection workers

Layer: Application. Context: Coordinates domain state and infrastructure to complete application use cases.

DetectionService

Orchestrates object detection operations and detection engine management.

Responsibilities

• Create and manage DetectionEngine instances (one per session, per model).
• Execute single-frame detection via detect_current_frame().
• Start background detection workers for processing entire sessions.
• Apply confidence and label filtering to detection results.
• Store detection results in session data layer.
• Handle model switching and clearing detection layers.

Note

Does not directly create DetectionWorker (delegates to external code). Manages DetectionEngine lifecycle (creation, model updates). Filters raw YOLO detections by confidence and chosen labels. Stores filtered results in DataLayer.B (detection layer).

detect_current_frame() — Single-frame detection (blocking, for UI immediate use). create_or_update_detection_engine_from_model_name_for_session_id() — Model management. start_detection_worker_for_session_id() — Background detection (threading).

PARAMETER	DESCRIPTION
app	Application adapted for session access, model management, detection, and worker coordination. TYPE: Application

__init__(app: Application) -> None

detect_current_frame_and_populate_layer_b_at_current_index(s_id: SessionId) -> None

Run selected current-frame detection via the execution collaborator.

PARAMETER	DESCRIPTION
s_id	Session whose current frame should be detected. TYPE: SessionId

Raises: Exception: Propagates detection or storage failures.

Note

The ...layer_b... name is a backward-compatible legacy API surface. Label/confidence-filtered output is stored in Layer A and cloned into Layer B while preserving manual annotations. The compatibility frame seeder can still populate Layer B for older data.

Example

service.detect_current_frame_and_populate_layer_b_at_current_index(s_id)

on_detection_batch_ready(s_id: SessionId, batch: dict[int, list[DetectionResult]], batch_processing_time: float) -> None

Runs on the Main Thread safely. Receives batches of detections from the worker and pushes them into the Domain Session.

on_detection_progress_updated(processed_frames: int, total_frames: int, eta_msecs: float | None) -> None

Forward detection progress updates to the UI-facing callback.

PARAMETER	DESCRIPTION
processed_frames	Number of frames processed so far. TYPE: int
total_frames	Total number of frames in the detection job. TYPE: int
eta_msecs	Estimated remaining time in milliseconds, if available. TYPE: float | None

set_detection_model_for_session_id(s_id: SessionId, model_name: str, keep_manual: bool = False) -> None

Sets the detection model for a given session and updates its settings.

start_detection_worker_for_session_id(s_id: SessionId) -> None

Starts a background DetectionWorker .