Filter

Filter wrapper.

This module provides a minimal wrapper around OIDN filters.

Filters are created from a :class:pyoidn.device.Device, then configured by setting images and parameters, and finally committed/executed.

Typical usage:

.. code-block:: python

    import numpy as np
    import pyoidn

    result = np.zeros_like(color, dtype=np.float32)
    with pyoidn.Device() as device:
            device.commit()
            with pyoidn.Filter(device, pyoidn.OIDN_FILTER_TYPE_RT) as flt:
                    flt.set_image(pyoidn.OIDN_IMAGE_COLOR, color, pyoidn.OIDN_FORMAT_FLOAT3)
                    flt.set_image(pyoidn.OIDN_IMAGE_OUTPUT, result, pyoidn.OIDN_FORMAT_FLOAT3)
                    flt.commit()
                    flt.execute()

    assert device.get_error() is None

Notes

• Image inputs can be passed either as NumPy arrays (shared host memory) or as :class:pyoidn.buffer.Buffer instances (OIDN buffers).
• Error handling: check :meth:pyoidn.device.Device.get_error after operations.
• Refer to https://github.com/RenderKit/oidn?tab=readme-ov-file#filters for the OIDN property settings.

Filter

Filter(device: Device, filter_type: str = OIDN_FILTER_TYPE_RT)

OIDN filter.

Parameters:

Name	Type	Description	Default
device	Device	The owning :class:pyoidn.device.Device.	required
filter_type	str	Filter type string (e.g. OIDN_FILTER_TYPE_RT). See OIDN docs for supported filter types.	OIDN_FILTER_TYPE_RT

Create a filter.

__enter__

__enter__() -> Filter

Enter a context manager.

Returns:

Type	Description
Filter	This filter.

__exit__

__exit__(exc_type, exc, tb)

Exit a context manager and release the filter.

commit

commit()

Commit filter parameters and images.

execute

execute()

Execute the filter synchronously.

execute_async

execute_async()

Execute the filter asynchronously.

Synchronize via :meth:pyoidn.device.Device.wait.

get_bool

get_bool(name: str) -> bool

Get a boolean filter parameter.

get_float

get_float(name: str) -> float

Get a float filter parameter.

get_int

get_int(name: str) -> int

Get an integer filter parameter.

release

release()

Release the underlying OIDN filter handle.

set_bool

set_bool(name: str, value: bool)

Set a boolean filter parameter.

set_float

set_float(name: str, value: float)

Set a float filter parameter.

set_image

set_image(name: str, data: Any, data_format: int, width: int = -1, height: int = -1, byte_offset: int = 0, pixel_byte_stride: int = 0, row_byte_stride: int = 0) -> None

Set an input/output image.

When data is a NumPy array, this calls OIDN's shared-image API (oidnSetSharedFilterImage). When data is a :class:pyoidn.buffer.Buffer, it uses the buffer-based API (oidnSetFilterImage).

Parameters:

Name	Type	Description	Default
name	str	Image slot name. Common values are: OIDN_IMAGE_COLOR, OIDN_IMAGE_ALBEDO, OIDN_IMAGE_NORMAL, OIDN_IMAGE_OUTPUT.	required
data	Any	NumPy array (host memory), :class:pyoidn.buffer.Buffer or torch.Tensor (if PyTorch is available). The array/tensor must be contiguous.	required
data_format	int	One of OIDN_FORMAT_* (e.g. OIDN_FORMAT_FLOAT3).	required
width	int	Image width. If data is a NumPy array and this is negative, width is derived from data.shape[1].	-1
height	int	Image height. If data is a NumPy array and this is negative, height is derived from data.shape[0].	-1
byte_offset	int	Byte offset into the buffer/pointer.	0
pixel_byte_stride	int	Bytes between consecutive pixels. Use 0 to let OIDN assume tightly packed pixels.	0
row_byte_stride	int	Bytes between consecutive rows. Use 0 to let OIDN assume tightly packed rows.	0

set_int

set_int(name: str, value: int)

Set an integer filter parameter.

set_quality

set_quality(quality: int)

Set the filter quality level.

Parameters:

Name	Type	Description	Default
quality	int	One of OIDN_QUALITY_*.	required

set_value

set_value(k: str, value)

Convenience setter that dispatches based on Python type.

unset_image

unset_image(name: str)

Unset an input/output image.

Parameters:

Name	Type	Description	Default
name	str	Image slot name.	required