benchmarks.hota.FramePrediction
class FramePrediction:One frame of model output: per-instance thing masks + semantics.
Members
method
property
- categories
- centroids
- embeddings
- heightMask height ``H``.
- masks
- num_instancesNumber of predicted instances ``M``.
- scores
- widthMask width ``W``.
Source: unitrack/benchmarks/hota/types.py:12