Callbacks

frogbox.callbacks.image_logger

ImageLogger

ImageLogger(
    split="test",
    log_label="images",
    model_key="model",
    resize_to_fit=True,
    interpolation="nearest",
    num_cols=None,
    show_progress=False,
    forward=None,
)

Bases: Callback

Callback for logging images.

Parameters:

• split

  (str, default: 'test' ) –

  Dataset split to evaluate on. Defaults to "test".

• log_label

  (str, default: 'images' ) –

  Label to log images under in Weights & Biases.

• model_key

  (str, default: 'model' ) –

  Pipeline model to use for inference.

• resize_to_fit

  (bool, default: True ) –

  If true smaller images are resized to fit canvas.

• interpolation

  (InterpolationMode, default: 'nearest' ) –

  Interpolation to use for resizing images.

• forward

  (callable, default: None ) –

  Function that arguments x, y and model and tuple of images to log. Returns (x, model(x), y) if not provided.

• show_progress

  (bool, default: False ) –

  Show progress bar.

• num_cols

  (int, default: None ) –

  Number of columns in image grid. Defaults to number of elements in returned tuple.