predict_split

Predictor.predict_split(split, img_ids=None, reps=('latent', 'decoder'), **kwargs)

Predict data from train/val/test split of dataset that the model was trained with.

Saves results in experiment_dir/exp_name/split.

Parameters

• split (str) – Data split to predict. One of train, val, test, val_imgs, test_imgs.

• img_ids (Union[ndarray, List[int], int, None]) – Object ids or number of objects that should be predicted (only for val_imgs and test_imgs).

• reps (Iterable[str]) – Representations to predict. See Predictor.get_representation().

Return type

None