spdl.io.save_image

save_image(path: str | Path, data: object, pix_fmt: str = 'rgb24', **kwargs)[source]

Save the given image tensor.

Note

The current implementation only supports writing data without changing the pixel format. If you need to apply preprocessing including pixel format conversion or rescaling, then construct such pipeline using FitlerGraph and Muxer.

Parameters:

  • path – The output path

  • data –

    The data to write. The shape must be one of the following and must match the value of pix_fmt.

    • "rgb24", "bgr24": (height, width, channel==3)

    • "gray", "gray16be": (height, width)

    • "yuv444p": (channel==3, height, width)

    If writing "gray16be", the dtype must be 16-bit signed integer. Otherwise it must be 8-bit unsigned integer.

  • pix_fmt – The pixel format used to write data. Supported values are, "rgb24", "bgr24", "gray", "gray16be", "gray16le", "gray", "yuv444p".

  • **kwargs – Extra arguments passed to video_encode_config().