Merge pull request #197 from RistoAle97/main

Fixed typos

Author: chaofengc

docs/Dataset_Preparation.md

@@ -17,7 +17,7 @@ The following datasets can be loaded with the current codes after downloaded (se
 | BAPPS      | *2AFC*      | SPAQ             | *Mobile*           |
 | PieAPP     | *2AFC*      | AVA              | *Aesthetic*        |
 | KADID-10k  |             | KonIQ-10k(++)    |                    |
-| LIVEM      |             | LIVEChallange    |                    |
+| LIVEM      |             | LIVEChallenge    |                    |
 | LIVE       |             | [PIQ2023](https://github.com/DXOMARK-Research/PIQ2023)| Portrait dataset   |
 | TID2013    |             | [GFIQA](http://database.mmsp-kn.de/gfiqa-20k-database.html)| Face IQA Dataset   |
 | TID2008    |             |                  |                    |

pyiqa/archs/brisque_arch.py

@@ -16,7 +16,7 @@
 from pyiqa.utils.color_util import to_y_channel
 from pyiqa.matlab_utils import imresize
 from pyiqa.matlab_utils.nss_feature import compute_nss_features
-from .func_util import estimate_ggd_param, estimate_aggd_param, normalize_img_with_guass
+from .func_util import estimate_ggd_param, estimate_aggd_param, normalize_img_with_gauss
 from pyiqa.utils.download_util import load_file_from_url
 from pyiqa.utils.registry import ARCH_REGISTRY
 
@@ -65,7 +65,7 @@ def brisque(x: torch.Tensor,
     num_of_scales = 2
     for _ in range(num_of_scales):
         if version == 'matlab': 
-            xnorm = normalize_img_with_guass(x, kernel_size, kernel_sigma, padding='replicate')
+            xnorm = normalize_img_with_gauss(x, kernel_size, kernel_sigma, padding='replicate')
             features.append(compute_nss_features(xnorm))
         elif version == 'original':
             features.append(natural_scene_statistics(x, kernel_size, kernel_sigma))
@@ -88,7 +88,7 @@ def brisque(x: torch.Tensor,
 
 def natural_scene_statistics(luma: torch.Tensor, kernel_size: int = 7, sigma: float = 7. / 6) -> torch.Tensor:
 
-    luma_nrmlzd = normalize_img_with_guass(luma, kernel_size, sigma, padding='same')
+    luma_nrmlzd = normalize_img_with_gauss(luma, kernel_size, sigma, padding='same')
     alpha, sigma = estimate_ggd_param(luma_nrmlzd)
     features = [alpha, sigma.pow(2)]
 

pyiqa/archs/ckdn_arch.py

@@ -4,7 +4,7 @@
 "Learning conditional knowledge distillation for degraded-reference image quality assessment." In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 10242-10251. 2021.
 
 Ref url: https://github.com/researchmm/CKDN.
-Re-implmented by: Chaofeng Chen (https://github.com/chaofengc)
+Re-implemented by: Chaofeng Chen (https://github.com/chaofengc)
 
 """
 

pyiqa/archs/clipiqa_arch.py

@@ -5,7 +5,7 @@
 AAAI 2023.
 
 Ref url: https://github.com/IceClear/CLIP-IQA
-Re-implmented by: Chaofeng Chen (https://github.com/chaofengc) with the following modification:
+Re-implemented by: Chaofeng Chen (https://github.com/chaofengc) with the following modification:
     - We assemble multiple prompts to improve the results of clipiqa model.
 
 """
@@ -60,7 +60,7 @@ def __init__(self, clip_model, n_ctx=16) -> None:
         self.register_buffer("token_prefix", init_embedding[:, :1, :])  # SOS
         self.register_buffer("token_suffix", init_embedding[:, 1 + n_ctx:, :])  # CLS, EOS
 
-    def get_prompts_with_middel_class(self,):
+    def get_prompts_with_middle_class(self,):
 
         ctx = self.ctx.to(self.token_prefix)
         if ctx.dim() == 2:
@@ -90,8 +90,8 @@ def get_prompts_with_middel_class(self,):
         return prompts
 
     def forward(self, clip_model):
-        prompts = self.get_prompts_with_middel_class()
-        # self.get_prompts_with_middel_class
+        prompts = self.get_prompts_with_middle_class()
+        # self.get_prompts_with_middle_class
         x = prompts + clip_model.positional_embedding.type(clip_model.dtype)
         x = x.permute(1, 0, 2)  # NLD -> LND
         x = clip_model.transformer(x)

pyiqa/archs/clipscore_arch.py

@@ -9,7 +9,7 @@
     }
 
 Reference url: https://github.com/jmhessel/clipscore
-Re-implmented by: Chaofeng Chen (https://github.com/chaofengc) 
+Re-implemented by: Chaofeng Chen (https://github.com/chaofengc) 
 """
 import torch
 import torch.nn as nn

pyiqa/archs/cnniqa_arch.py

@@ -28,7 +28,7 @@ class CNNIQA(nn.Module):
     r"""CNNIQA model.
     Args:
         - ker_size (int): Kernel size.
-        - n_kers (int): Number of kernals.
+        - n_kers (int): Number of kernels.
         - n1_nodes (int): Number of n1 nodes.
         - n2_nodes (int): Number of n2 nodes.
         - pretrained_model_path (String): Pretrained model path.

pyiqa/archs/dists_arch.py

@@ -55,7 +55,7 @@ class DISTS(torch.nn.Module):
     '''
 
     def __init__(self, pretrained=True, pretrained_model_path=None, **kwargs):
-        """Refer to offical code https://github.com/dingkeyan93/DISTS
+        """Refer to official code https://github.com/dingkeyan93/DISTS
         """
         super(DISTS, self).__init__()
         vgg_pretrained_features = models.vgg16(weights='IMAGENET1K_V1').features

pyiqa/archs/fid_arch.py

@@ -316,7 +316,7 @@ def forward(self,
 
         # compute fid of a folder
         elif fdir1 is not None and fdir2 is None:
-            assert dataset_name is not None, "When fdir2 is not provided, the reference dataset_name should be specified to calcualte fid score."
+            assert dataset_name is not None, "When fdir2 is not provided, the reference dataset_name should be specified to calculate fid score."
             if verbose:
                 print(f"compute FID of a folder with {dataset_name}-{mode}-{dataset_split}-{dataset_res} statistics")
             fbname1 = os.path.basename(fdir1)

pyiqa/archs/fsim_arch.py

@@ -58,7 +58,7 @@ def fsim(x: torch.Tensor,
             threshold  point, below which phase congruency values get penalized.
 
     Returns:
-        - Index of similarity betwen two images. Usually in [0, 1] interval.
+        - Index of similarity between two images. Usually in [0, 1] interval.
         Can be bigger than 1 for predicted :math:`x` images with higher contrast than the original ones.
     References:
         L. Zhang, L. Zhang, X. Mou and D. Zhang, "FSIM: A Feature Similarity Index for Image Quality Assessment,"
@@ -369,7 +369,7 @@ def _lowpassfilter(size: Tuple[int, int], cutoff: float, n: int) -> torch.Tensor
     r"""
     Constructs a low-pass Butterworth filter.
     Args:
-        size: Tuple with heigth and width of filter to construct
+        size: Tuple with height and width of filter to construct
         cutoff: Cutoff frequency of the filter in (0, 0.5()
         n: Filter order. Higher `n` means sharper transition.
             Note that `n` is doubled so that it is always an even integer.

pyiqa/archs/func_util.py

@@ -83,7 +83,7 @@ def diff_round(x: torch.Tensor) -> torch.Tensor:
     return x - x.detach() + x.round()
 
 
-def normalize_img_with_guass(
+def normalize_img_with_gauss(
     img: torch.Tensor,
     kernel_size: int = 7,
     sigma: float = 7.0 / 6,
@@ -128,7 +128,7 @@ def similarity_map(
         map_x: Tensor with map to be compared
         map_y: Tensor with map to be compared
         constant: Used for numerical stability
-        alpha: Masking coefficient. Substracts - `alpha` * map_x * map_y from denominator and nominator
+        alpha: Masking coefficient. Subtracts - `alpha` * map_x * map_y from denominator and nominator
     """
     return (2.0 * map_x * map_y - alpha * map_x * map_y + constant) / (
         map_x**2 + map_y**2 - alpha * map_x * map_y + constant + EPS

pyiqa/archs/inception.py

@@ -24,7 +24,7 @@ class InceptionV3(nn.Module):
     # Maps feature dimensionality to their output blocks indices
     BLOCK_INDEX_BY_DIM = {
         64: 0,   # First max pooling features
-        192: 1,  # Second max pooling featurs
+        192: 1,  # Second max pooling features
         768: 2,  # Pre-aux classifier features
         2048: 3  # Final average pooling features
     }
@@ -176,7 +176,7 @@ def forward(self, inp, resize_input=False, normalize_input=False):
 
 def _inception_v3(*args, **kwargs):
     """Wraps `torchvision.models.inception_v3`
-    Skips default weight inititialization if supported by torchvision version.
+    Skips default weight initialization if supported by torchvision version.
     See https://github.com/mseitzer/pytorch-fid/issues/28.
     """
     try:

pyiqa/archs/iqt_arch.py

@@ -9,7 +9,7 @@
 }
 
 Ref url: https://github.com/anse3832/IQT
-Re-implmented by: Chaofeng Chen (https://github.com/chaofengc)
+Re-implemented by: Chaofeng Chen (https://github.com/chaofengc)
 
 """
 

pyiqa/archs/lpips_arch.py

@@ -139,7 +139,7 @@ def forward(self, in1, in0, retPerLayer=False, normalize=True):
         Args:
             in1: An input tensor. Shape :math:`(N, C, H, W)`.
             in0: A reference tensor. Shape :math:`(N, C, H, W)`.
-            retPerLayer (Boolean): return result contains ressult of
+            retPerLayer (Boolean): return result contains result of
                 each layer or not. Default: False.
             normalize (Boolean): Whether to normalize image data range
                 in [0,1] to [-1,1]. Default: True.

pyiqa/archs/mad_arch.py

@@ -5,7 +5,7 @@
 Modified by: Jiadi Mo (https://github.com/JiadiMo)
 
 Note:
-    Offical matlab code is not available;
+    Official matlab code is not available;
     Pytorch version >= 1.8.0;
 """
 

pyiqa/archs/maniqa_swin.py

@@ -162,7 +162,7 @@ class SwinBlock(nn.Module):
 
     Args:
         dim (int): Number of input channels.
-        input_resolution (tuple[int]): Input resulotion.
+        input_resolution (tuple[int]): Input resolution.
         num_heads (int): Number of attention heads.
         window_size (int): Window size.
         shift_size (int): Shift size for SW-MSA.

pyiqa/archs/niqe_arch.py

@@ -21,7 +21,7 @@
 from pyiqa.utils.color_util import to_y_channel
 from pyiqa.utils.download_util import load_file_from_url
 from pyiqa.matlab_utils import imresize, fspecial, conv2d, imfilter, fitweibull, nancov, nanmean, blockproc
-from .func_util import estimate_aggd_param, normalize_img_with_guass, diff_round
+from .func_util import estimate_aggd_param, normalize_img_with_gauss, diff_round
 from pyiqa.archs.fsim_arch import _construct_filters
 from pyiqa.utils.registry import ARCH_REGISTRY
 
@@ -119,7 +119,7 @@ def niqe(img: torch.Tensor,
 
     distparam = []  # dist param is actually the multiscale features
     for scale in (1, 2):  # perform on two scales (1, 2)
-        img_normalized = normalize_img_with_guass(img, padding='replicate')
+        img_normalized = normalize_img_with_gauss(img, padding='replicate')
 
         distparam.append(blockproc(img_normalized, [block_size_h // scale, block_size_w // scale], fun=compute_feature))
 
@@ -261,7 +261,7 @@ def ilniqe(img: torch.Tensor,
 
     distparam = []  # dist param is actually the multiscale features
     for scale in (1, 2):  # perform on two scales (1, 2)
-        struct_dis = normalize_img_with_guass(O_img[:, [2]], kernel_size=5, sigma=5. / 6, padding='replicate')
+        struct_dis = normalize_img_with_gauss(O_img[:, [2]], kernel_size=5, sigma=5. / 6, padding='replicate')
 
         dx, dy = gauDerivative(sigmaForGauDerivative / (scale**scaleFactorForGaussianDer), device=img)
 

pyiqa/archs/nlpd_arch.py

@@ -149,7 +149,7 @@ def nlpd(self, x1, x2):
             x1: An input tensor. Shape :math:`(N, C, H, W)`.
             x2: A target tensor. Shape :math:`(N, C, H, W)`.
         Returns:
-            Index of similarity betwen two images. Usually in [0, 1] interval.
+            Index of similarity between two images. Usually in [0, 1] interval.
         """
         assert (self.test_y_channel and self.channels == 1) or (
             not self.test_y_channel and self.channels == 3), 'Number of channel and convert to YCBCR should be match'

pyiqa/archs/nrqm_arch.py

@@ -26,7 +26,7 @@
 default_model_urls = {'url': 'https://github.com/chaofengc/IQA-PyTorch/releases/download/v0.1-weights/NRQM_model.mat'}
 
 
-def get_guass_pyramid(x: Tensor, scale: int = 2):
+def get_gauss_pyramid(x: Tensor, scale: int = 2):
     r"""Get gaussian pyramid images with gaussian kernel.
     """
     pyr = [x]
@@ -327,7 +327,7 @@ def nrqm(
     # crop image
     b, c, h, w = img.shape
     img = img.double()
-    img_pyr = get_guass_pyramid(img / 255.)
+    img_pyr = get_gauss_pyramid(img / 255.)
 
     # DCT features
     f1 = []

pyiqa/archs/piqe_arch.py

@@ -15,7 +15,7 @@
 
 from pyiqa.utils.color_util import to_y_channel
 from pyiqa.matlab_utils import symm_pad
-from pyiqa.archs.func_util import normalize_img_with_guass
+from pyiqa.archs.func_util import normalize_img_with_gauss
 from pyiqa.utils.registry import ARCH_REGISTRY
 
 
@@ -52,7 +52,7 @@ def piqe(
     new_height, new_width = img.shape[2], img.shape[3]
 
     # Normalize image to zero mean and ~unit std
-    img_normalized = normalize_img_with_guass(img, padding='replicate')
+    img_normalized = normalize_img_with_gauss(img, padding='replicate')
 
     # Create blocks
     blocks = img_normalized.unfold(2, block_size, block_size).unfold(3, block_size, block_size)

pyiqa/archs/psnr_arch.py

@@ -28,7 +28,7 @@ def psnr(x, y, test_y_channel=False, data_range=1.0, eps=1e-8, color_space='yiq'
         - data_range: Maximum value range of images (default 1.0).
 
     Returns:
-        PSNR Index of similarity betwen two images.
+        PSNR Index of similarity between two images.
     """
 
     if (x.shape[1] == 3) and test_y_channel:
@@ -50,7 +50,7 @@ class PSNR(nn.Module):
         - test_y_channel (Boolean): Convert RGB image to YCbCr format and computes PSNR
             only on luminance channel if `True`. Compute on all 3 channels otherwise.
         - kwargs: other parameters, including
-            - data_range: maximun numeric value
+            - data_range: maximum numeric value
             - eps: small constant for numeric stability
     Return:
         score (torch.Tensor): (B, 1)

pyiqa/archs/ssim_arch.py

@@ -8,11 +8,11 @@
 Modified by: Jiadi Mo (https://github.com/JiadiMo)
 
 Refer to:
-    - Offical SSIM matlab code from https://www.cns.nyu.edu/~lcv/ssim/;
+    - Official SSIM matlab code from https://www.cns.nyu.edu/~lcv/ssim/;
     - PIQ from https://github.com/photosynthesis-team/piq;
     - BasicSR from https://github.com/xinntao/BasicSR/blob/master/basicsr/metrics/psnr_ssim.py;
-    - Offical MS-SSIM matlab code from https://ece.uwaterloo.ca/~z70wang/research/iwssim/msssim.zip;
-    - Offical CW-SSIM matlab code from
+    - Official MS-SSIM matlab code from https://ece.uwaterloo.ca/~z70wang/research/iwssim/msssim.zip;
+    - Official CW-SSIM matlab code from
     https://www.mathworks.com/matlabcentral/mlc-downloads/downloads/submissions/43017/versions/1/download/zip;
 
 """
@@ -118,7 +118,7 @@ def ms_ssim(X, Y, win=None, data_range=1., downsample=False, test_y_channel=True
         test_y_channel: Boolean, whether to use y channel on ycbcr.
         is_prod: Boolean, calculate product or sum between mcs and weight.
     Returns:
-        Index of similarity betwen two images. Usually in [0, 1] interval.
+        Index of similarity between two images. Usually in [0, 1] interval.
     """
     if not X.shape == Y.shape:
         raise ValueError('Input images must have the same dimensions.')
@@ -248,7 +248,7 @@ def cw_ssim(self, x, y, test_y_channel):
             y: A target tensor. Shape :math:`(N, C, H, W)`.
             test_y_channel: Boolean, whether to use y channel on ycbcr.
         Returns:
-            Index of similarity betwen two images. Usually in [0, 1] interval.
+            Index of similarity between two images. Usually in [0, 1] interval.
         """
         # Whether calculate on y channel of ycbcr
         if test_y_channel and x.shape[1] == 3:

pyiqa/archs/stlpips_arch.py

@@ -123,7 +123,7 @@ def forward(self, in0, in1, retPerLayer=False, normalize=True):
         Args:
             in1: An input tensor. Shape :math:`(N, C, H, W)`.
             in0: A reference tensor. Shape :math:`(N, C, H, W)`.
-            retPerLayer (Boolean): return result contains ressult of
+            retPerLayer (Boolean): return result contains result of
                 each layer or not. Default: False.
             normalize (Boolean): Whether to normalize image data range
                 in [0,1] to [-1,1]. Default: True.

pyiqa/archs/topiq_swin.py

@@ -270,7 +270,7 @@ class SwinTransformerBlock(nn.Module):
 
     Args:
         dim (int): Number of input channels.
-        input_resolution (tuple[int]): Input resulotion.
+        input_resolution (tuple[int]): Input resolution.
         window_size (int): Window size.
         num_heads (int): Number of attention heads.
         head_dim (int): Enforce the number of channels per head