Skip to main content
Version: v1.7

Pipeline Builder API

important

Preview Core operators (detection, classification, pose, segmentation, tracking) are stable. Cascade (op.for_each, op.crop_roi) and streaming APIs are still in development.

The Pipeline Builder API is the Pythonic interface for building ML inference pipelines on the Axelera Metis AIPU. It replaces YAML-based pipeline configuration with composable Python operators.

Note:

The Operator classes each have a snake_case alias. e.g.:

  • op.Seq can also be used as op.seq
  • op.CenterCrop can also be used as op.center_crop
  • etc.

Getting started​

Operator Reference​

  • Transforms — Image preprocessing (resize, letterbox, normalize, color convert)
  • Postprocess — Decode raw model output, NMS, coordinate transform, filter
  • Results — Convert arrays to typed objects (DetectedObject, PoseObject, etc.)
  • Inference — Model loading (op.load, op.onnx_model)
  • Tracker — Multi-object tracking (ByteTrack, OC-SORT, SORT, TrackTrack)
  • Combinators — Pipeline composition (op.seq, op.par, op.for_each)

Running Pipelines​

  • Sources — Video/stream input (cv.create_source, VideoSource, Image)
  • Scheduler — Device selection and execution backing pipeline.stream / pipeline.batch

Type Reference​

  • Types — Data types: BBox, DetectedObject, PoseObject, SegmentedObject, TrackedObject, Classification