Image transformer

`ImageTransformer`

Replaces each declared image column of a DataFrame by numeric features.

ImageTransformer(
    image_features_indices: Sequence[int],
    *,
    n_components: int = DEFAULT_N_COMPONENTS,
    device: Any = "auto",
    batch_size: int = DEFAULT_BATCH_SIZE,
)

Parameters

Parameter	Type	Default	Description
`image_features_indices`	`Sequence[int]`	Required	Positions of the image columns in `X`. Such a column holds one image per cell, as a base64 string, the image file’s path or bytes, or a PIL image, so its dtype is object or string.
`n_components`	`int`	`DEFAULT_N_COMPONENTS`	Features an image column is expanded into.
`device`	`Any`	`"auto"`	Where the encoder runs, as TabPFN’s `device` argument.
`batch_size`	`int`	`DEFAULT_BATCH_SIZE`	Images per encoder forward pass.

Attributes

Attribute	Type	Description
`reducers_`	`dict[int, Pipeline]`	Input position -> the column’s fitted scaler and PCA.
`feature_names_in_`	`np.ndarray`	Labels of the frame `fit` saw, as strings.

`ImageTransformer.fit`

View source

Fit one scaler and PCA per declared column; y is ignored.

ImageTransformer.fit(
    X: pd.DataFrame,
    y: Any = None,
) -> ImageTransformer

Parameters

Parameter	Type	Default	Description
`X`	`pd.DataFrame`	Required	—
`y`	`Any`	`None`	—

Returns

Type	Description
`ImageTransformer`	—

`ImageTransformer.fit_transform`

View source

Fit every declared column and return the expanded frame.

ImageTransformer.fit_transform(
    X: pd.DataFrame,
    y: Any = None,
    **fit_params: Any,
) -> pd.DataFrame

Parameters

Parameter	Type	Default	Description
`X`	`pd.DataFrame`	Required	—
`y`	`Any`	`None`	—
`**fit_params`	`Any`	—	—

Returns

Type	Description
`pd.DataFrame`	—

`ImageTransformer.get_feature_names_out`

View source

The kept columns’ labels, then the image features’, as an object array.

ImageTransformer.get_feature_names_out(
    input_features: Any = None,
) -> np.ndarray

Parameters

Parameter	Type	Default	Description
`input_features`	`Any`	`None`	—

Returns

Type	Description
`np.ndarray`	—

`ImageTransformer.output_indices`

View source

Where the input positions indices sit once the image columns moved.

ImageTransformer.output_indices(
    indices: Sequence[int] | None,
    *,
    n_columns: int,
) -> list[int] | None

Parameters

Parameter	Type	Default	Description
`indices`	`Sequence[int] \| None`	Required	Positions in a frame of `n_columns` columns; `None` stays `None`.
`n_columns`	`int`	Required	The width of that frame.

Raises ValueError If one of them is a declared image column. Returns

Type	Description
`list[int] \| None`	—

`ImageTransformer.transform`

View source

Replace the declared columns of X by the features fit at training.

ImageTransformer.transform(
    X: pd.DataFrame,
) -> pd.DataFrame

Parameters

Parameter	Type	Default	Description
`X`	`pd.DataFrame`	Required	—

Returns

Type	Description
`pd.DataFrame`	—

`GatedEncoderError`

View source

The encoder’s weights are gated on the Hub and this process may not read them.

GatedEncoderError()

`image_to_bytes`

View source

A PIL image encoded as an image file, the form a DataFrame cell takes.

image_to_bytes(
    image: Image,
    format: str = "PNG",
) -> bytes

Parameters

Parameter	Type	Default	Description
`image`	`Image`	Required	—
`format`	`str`	`"PNG"`	—

Returns

Type	Description
`bytes`	—

Reference

tabpfn

tabpfn-client

tabpfn-extensions

REST API

`ImageTransformer`

`ImageTransformer.fit`

`ImageTransformer.fit_transform`

`ImageTransformer.get_feature_names_out`

`ImageTransformer.output_indices`

`ImageTransformer.transform`

`GatedEncoderError`

`image_to_bytes`

​ImageTransformer

​ImageTransformer.fit

​ImageTransformer.fit_transform

​ImageTransformer.get_feature_names_out

​ImageTransformer.output_indices

​ImageTransformer.transform

​GatedEncoderError

​image_to_bytes

`ImageTransformer`

`ImageTransformer.fit`

`ImageTransformer.fit_transform`

`ImageTransformer.get_feature_names_out`

`ImageTransformer.output_indices`

`ImageTransformer.transform`

`GatedEncoderError`

`image_to_bytes`