HINet

Warning

This is very large model and the weight file takes approximately 400MB on the disk.

Original image and denoised image

BiWAKO.HINet

HINet for denoising image.

Attributes:

Name	Type	Description
model	InferenceSession	ONNX model.
input_name	str	Name of input node.
input_shape	tuple	Shape of input node.

__init__(self, model='denoise_320_480') special

Initialize HINet.

Parameters:

Name	Type	Description	Default
model	str	Choise of model. Weight file is automatically downloaded to the current directory at the first time. Defaults to "denoise_320_480.onnx".	'denoise_320_480'

predict(self, image)

Return denoised image.

Parameters:

Name	Type	Description	Default
image	Image	Image to be denoised in str or cv2 format.	required

Returns:

Type	Description
np.ndarray	Denoised image array containing two images for different denoising methods.

render(self, prediction, image=None, output_type=0, output_shape=None)

Return the denoised image in original image size.

Parameters:

Name	Type	Description	Default
prediction	np.ndarray	Return value of predict().	required
image	Image	Image to be processed in str or cv2 format. Defaults to None.	None
output_type	Literal[0, 1]	Choice of denoising method either 0 or 1. Defaults to 0.	0
output_shape	Optional[Tuple[int, int]]	Optional tuple of int to resize the return image. Defaults to None.	None

Exceptions:

Type	Description
ValueError	If none of the original image or image size is given.
ValueError	If output_type is not 0 or 1.

Returns:

Type	Description
np.ndarray	Denoised image in cv2 format.