nerfbaselines.evaluation

class nerfbaselines.evaluation.DefaultEvaluationProtocol

Bases: EvaluationProtocol

accumulate_metrics(metrics: Iterable[Dict[str, float | int]]) → Dict[str, float | int]

evaluate(predictions: Dict[str, ndarray], dataset: Dataset) → Dict[str, float | int]

get_name()

render(method: Method, dataset: Dataset, *, options=None) → Dict[str, ndarray]

class nerfbaselines.evaluation.NerfEvaluationProtocol

Bases: DefaultEvaluationProtocol

nerfbaselines.evaluation.build_evaluation_protocol(id: str) → EvaluationProtocol

nerfbaselines.evaluation.compute_metrics(pred: np.ndarray, gt: np.ndarray, *, reduce: Literal[True] = True, run_lpips_vgg: bool = False) → Dict[str, float]

nerfbaselines.evaluation.compute_metrics(pred: np.ndarray, gt: np.ndarray, *, reduce: Literal[False], run_lpips_vgg: bool = False) → Dict[str, np.ndarray]

nerfbaselines.evaluation.evaluate(predictions: str, output: str, description: str = 'evaluating', evaluation_protocol: EvaluationProtocol | None = None)

Evaluate a set of predictions.

Parameters:

• predictions – Path to a directory containing the predictions.

• output – Path to a json file where the results will be written.

• description – Description of the evaluation, used for progress bar.

• evaluation_protocol – The evaluation protocol to use. If None, the protocol from info.json will be used.

Returns:

A dictionary containing the results.

nerfbaselines.evaluation.path_is_video(path: str) → bool

nerfbaselines.evaluation.render_all_images(method: Method, dataset: Dataset, output: str, description: str = 'rendering all images', nb_info: dict | None = None, evaluation_protocol: EvaluationProtocol | None = None) → Iterable[Dict[str, ndarray]]

nerfbaselines.evaluation.render_frames(method: Method, cameras: Cameras, output: str | Path, fps: float, embeddings: List[ndarray] | None = None, description: str = 'rendering frames', output_names: Tuple[str, ...] = ('color',), nb_info: dict | None = None) → None

nerfbaselines.evaluation.run_inside_eval_container(backend_name: str | None = None)

Ensures PyTorch is available to compute extra metrics (lpips)

nerfbaselines.evaluation.trajectory_get_cameras(trajectory: Trajectory) → Cameras

nerfbaselines.evaluation.trajectory_get_embeddings(method: Method, trajectory: Trajectory) → List[ndarray] | None

nerfbaselines.evaluation.with_supported_camera_models(supported_camera_models)