Specialized Blocks
Source: lib/workflow_blocks/specialized.py
Domain-specific detectors that wrap purpose-built models instead of generic YOLO/SAM.
Face Detection
Lightweight face detector. Used as a pre-filter before face-recognition or blur blocks.
Inputs: image
Outputs: faces: list[Box]
License Plate Detection
Detects license plates and (optionally) crops + OCRs them.
Inputs: image
Outputs: plates: list[Plate] where Plate = { box, text?, confidence }
Pose Estimation
Returns 17-keypoint COCO pose for each detected person.
Inputs: image
Outputs: poses: list[Pose]
OCR
General-purpose text recognition. Runs against a cropped region (usually downstream of a detector or crop block).
Inputs: image
Outputs: text: str, confidence: float
The full list of blocks updates as new ones are registered. Use GET /api/v1/workflows/blocks to fetch the live catalog with input/output schemas.