| from __future__ import annotations
|
|
|
| import sys
|
| from pathlib import Path
|
|
|
| _MODEL_DIR = Path(__file__).resolve().parent
|
| if str(_MODEL_DIR) not in sys.path:
|
| sys.path.insert(0, str(_MODEL_DIR))
|
|
|
| from mediapipe_mvp import has_mediapipe_artifact, load_mediapipe_mvp |
| from modeling import KslWordRecognizer, load_json, load_labels |
|
|
|
|
| def load_model(model_dir: str, device: str) -> KslWordRecognizer: |
| """Load the YOLO-assisted VideoMAE KSL word recognition pipeline.""" |
|
|
| root = Path(model_dir) |
| if has_mediapipe_artifact(root): |
| return load_mediapipe_mvp(root, device) |
|
|
| labels = load_labels(root / "labels.json") |
| preprocessor = load_json(root / "preprocessor_config.json", default={})
|
| config = load_json(root / "model_config.json", default={})
|
| return KslWordRecognizer(
|
| model_dir=root,
|
| labels=labels,
|
| preprocessor=preprocessor,
|
| config=config,
|
| device=device,
|
| )
|
|
|
|
|
| def predict(model: KslWordRecognizer, frames: list, timestamps_ms: list[int | None]) -> list[dict]: |
| """Return one word-level caption prediction per input RGB frame."""
|
|
|
| if len(frames) != len(timestamps_ms):
|
| raise ValueError("frames and timestamps_ms must have the same length.")
|
| return [model.predict_one(frame, timestamp_ms) for frame, timestamp_ms in zip(frames, timestamps_ms, strict=True)]
|
|
|
