Unitrack
type

benchmarks.hota.ModelAdapter

class ModelAdapter(Protocol):

A HuggingFace model that emits one ``FramePrediction`` per image.

Members

method

  • __call__Run inference on a single ``(H, W, 3)`` uint8 RGB frame.
  • loadMaterialize weights/processor on ``device`` (call once before use).

property

Source: unitrack/benchmarks/hota/protocols.py:17