File size: 10,030 Bytes
36c95ba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
import math
from typing import Dict, Optional

import torch
import torch.nn as nn

from kornia.filters.kernels import get_gaussian_kernel2d
from kornia.filters.sobel import SpatialGradient

from .laf import (
    ellipse_to_laf,
    extract_patches_from_pyramid,
    get_laf_scale,
    make_upright,
    raise_error_if_laf_is_not_valid,
    scale_laf,
)

urls: Dict[str, str] = {}
urls["affnet"] = "https://github.com/ducha-aiki/affnet/raw/master/pretrained/AffNet.pth"


class PatchAffineShapeEstimator(nn.Module):
    r"""Module, which estimates the second moment matrix of the patch gradients.

    The method determines the affine shape of the local feature as in :cite:`baumberg2000`.

    Args:
        patch_size: the input image patch size.
        eps: for safe division.
    """

    def __init__(self, patch_size: int = 19, eps: float = 1e-10):
        super().__init__()
        self.patch_size: int = patch_size
        self.gradient: nn.Module = SpatialGradient('sobel', 1)
        self.eps: float = eps
        sigma: float = float(self.patch_size) / math.sqrt(2.0)
        self.weighting: torch.Tensor = get_gaussian_kernel2d((self.patch_size, self.patch_size), (sigma, sigma), True)

    def __repr__(self):
        return self.__class__.__name__ + '(' 'patch_size=' + str(self.patch_size) + ', ' + 'eps=' + str(self.eps) + ')'

    def forward(self, patch: torch.Tensor) -> torch.Tensor:
        """Args:
            patch: (torch.Tensor) shape [Bx1xHxW]
        Returns:
            torch.Tensor: ellipse_shape shape [Bx1x3]"""
        if not isinstance(patch, torch.Tensor):
            raise TypeError(f"Input type is not a torch.Tensor. Got {type(patch)}")
        if not len(patch.shape) == 4:
            raise ValueError(f"Invalid input shape, we expect Bx1xHxW. Got: {patch.shape}")
        _, CH, W, H = patch.size()
        if (W != self.patch_size) or (H != self.patch_size) or (CH != 1):
            raise TypeError(
                "input shape should be must be [Bx1x{}x{}]. "
                "Got {}".format(self.patch_size, self.patch_size, patch.size())
            )
        self.weighting = self.weighting.to(patch.dtype).to(patch.device)
        grads: torch.Tensor = self.gradient(patch) * self.weighting
        # unpack the edges
        gx: torch.Tensor = grads[:, :, 0]
        gy: torch.Tensor = grads[:, :, 1]
        # abc == 1st axis, mixture, 2nd axis. Ellipse_shape is a 2nd moment matrix.
        ellipse_shape = torch.cat(
            [
                gx.pow(2).mean(dim=2).mean(dim=2, keepdim=True),
                (gx * gy).mean(dim=2).mean(dim=2, keepdim=True),
                gy.pow(2).mean(dim=2).mean(dim=2, keepdim=True),
            ],
            dim=2,
        )

        # Now lets detect degenerate cases: when 2 or 3 elements are close to zero (e.g. if patch is completely black
        bad_mask = ((ellipse_shape < self.eps).float().sum(dim=2, keepdim=True) >= 2).to(ellipse_shape.dtype)
        # We will replace degenerate shape with circular shapes.
        circular_shape = torch.tensor([1.0, 0.0, 1.0]).to(ellipse_shape.device).to(ellipse_shape.dtype).view(1, 1, 3)
        ellipse_shape = ellipse_shape * (1.0 - bad_mask) + circular_shape * bad_mask
        # normalization
        ellipse_shape = ellipse_shape / ellipse_shape.max(dim=2, keepdim=True)[0]
        return ellipse_shape


class LAFAffineShapeEstimator(nn.Module):
    """Module, which extracts patches using input images and local affine frames (LAFs).

    Then runs :class:`~kornia.feature.PatchAffineShapeEstimator` on patches to estimate LAFs shape.

    Then original LAF shape is replaced with estimated one. The original LAF orientation is not preserved,
    so it is recommended to first run LAFAffineShapeEstimator and then LAFOrienter.

    Args:
        patch_size: the input image patch size.
        affine_shape_detector: Patch affine shape estimator, :class:`~kornia.feature.PatchAffineShapeEstimator`.
    """  # pylint: disable

    def __init__(self, patch_size: int = 32, affine_shape_detector: Optional[nn.Module] = None) -> None:
        super().__init__()
        self.patch_size = patch_size
        self.affine_shape_detector = affine_shape_detector or PatchAffineShapeEstimator(self.patch_size)

    def __repr__(self):
        return (
            self.__class__.__name__ + '('
            'patch_size='
            + str(self.patch_size)
            + ', '
            + 'affine_shape_detector='
            + str(self.affine_shape_detector)
            + ')'
        )

    def forward(self, laf: torch.Tensor, img: torch.Tensor) -> torch.Tensor:
        """
        Args:
            laf: (torch.Tensor) shape [BxNx2x3]
            img: (torch.Tensor) shape [Bx1xHxW]

        Returns:
            torch.Tensor: laf_out shape [BxNx2x3]"""
        raise_error_if_laf_is_not_valid(laf)
        img_message: str = f"Invalid img shape, we expect BxCxHxW. Got: {img.shape}"
        if not isinstance(img, torch.Tensor):
            raise TypeError(f"img type is not a torch.Tensor. Got {type(img)}")
        if len(img.shape) != 4:
            raise ValueError(img_message)
        if laf.size(0) != img.size(0):
            raise ValueError(f"Batch size of laf and img should be the same. Got {img.size(0)}, {laf.size(0)}")
        B, N = laf.shape[:2]
        PS: int = self.patch_size
        patches: torch.Tensor = extract_patches_from_pyramid(img, make_upright(laf), PS, True).view(-1, 1, PS, PS)
        ellipse_shape: torch.Tensor = self.affine_shape_detector(patches)
        ellipses = torch.cat([laf.view(-1, 2, 3)[..., 2].unsqueeze(1), ellipse_shape], dim=2).view(B, N, 5)
        scale_orig = get_laf_scale(laf)
        laf_out = ellipse_to_laf(ellipses)
        ellipse_scale = get_laf_scale(laf_out)
        laf_out = scale_laf(laf_out, scale_orig / ellipse_scale)
        return laf_out


class LAFAffNetShapeEstimator(nn.Module):
    """Module, which extracts patches using input images and local affine frames (LAFs).

    Then runs AffNet on patches to estimate LAFs shape. This is based on the original code from paper
    "Repeatability Is Not Enough: Learning Discriminative Affine Regions via Discriminability"".
    See :cite:`AffNet2018` for more details.

    Then original LAF shape is replaced with estimated one. The original LAF orientation is not preserved,
    so it is recommended to first run LAFAffineShapeEstimator and then LAFOrienter.

    Args:
        pretrained: Download and set pretrained weights to the model.
    """

    def __init__(self, pretrained: bool = False):
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(1, 16, kernel_size=3, padding=1, bias=False),
            nn.BatchNorm2d(16, affine=False),
            nn.ReLU(),
            nn.Conv2d(16, 16, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(16, affine=False),
            nn.ReLU(),
            nn.Conv2d(16, 32, kernel_size=3, stride=2, padding=1, bias=False),
            nn.BatchNorm2d(32, affine=False),
            nn.ReLU(),
            nn.Conv2d(32, 32, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(32, affine=False),
            nn.ReLU(),
            nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1, bias=False),
            nn.BatchNorm2d(64, affine=False),
            nn.ReLU(),
            nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(64, affine=False),
            nn.ReLU(),
            nn.Dropout(0.25),
            nn.Conv2d(64, 3, kernel_size=8, stride=1, padding=0, bias=True),
            nn.Tanh(),
            nn.AdaptiveAvgPool2d(1),
        )
        self.patch_size = 32
        # use torch.hub to load pretrained model
        if pretrained:
            pretrained_dict = torch.hub.load_state_dict_from_url(
                urls['affnet'], map_location=lambda storage, loc: storage
            )
            self.load_state_dict(pretrained_dict['state_dict'], strict=False)
        self.eval()

    @staticmethod
    def _normalize_input(x: torch.Tensor, eps: float = 1e-6) -> torch.Tensor:
        """Utility function that normalizes the input by batch."""
        sp, mp = torch.std_mean(x, dim=(-3, -2, -1), keepdim=True)
        # WARNING: we need to .detach() input, otherwise the gradients produced by
        # the patches extractor with F.grid_sample are very noisy, making the detector
        # training totally unstable.
        return (x - mp.detach()) / (sp.detach() + eps)

    def forward(self, laf: torch.Tensor, img: torch.Tensor) -> torch.Tensor:
        """
        Args:
            laf: shape [BxNx2x3]
            img: shape [Bx1xHxW]

        Returns:
            laf_out shape [BxNx2x3]
        """
        raise_error_if_laf_is_not_valid(laf)
        img_message: str = f"Invalid img shape, we expect BxCxHxW. Got: {img.shape}"
        if not torch.is_tensor(img):
            raise TypeError(f"img type is not a torch.Tensor. Got {type(img)}")
        if len(img.shape) != 4:
            raise ValueError(img_message)
        if laf.size(0) != img.size(0):
            raise ValueError(f"Batch size of laf and img should be the same. Got {img.size(0)}, {laf.size(0)}")
        B, N = laf.shape[:2]
        PS: int = self.patch_size
        patches: torch.Tensor = extract_patches_from_pyramid(img, make_upright(laf), PS, True).view(-1, 1, PS, PS)
        xy = self.features(self._normalize_input(patches)).view(-1, 3)
        a1 = torch.cat([1.0 + xy[:, 0].reshape(-1, 1, 1), 0 * xy[:, 0].reshape(-1, 1, 1)], dim=2)
        a2 = torch.cat([xy[:, 1].reshape(-1, 1, 1), 1.0 + xy[:, 2].reshape(-1, 1, 1)], dim=2)
        new_laf_no_center = torch.cat([a1, a2], dim=1).reshape(B, N, 2, 2)
        new_laf = torch.cat([new_laf_no_center, laf[:, :, :, 2:3]], dim=3)
        scale_orig = get_laf_scale(laf)
        ellipse_scale = get_laf_scale(new_laf)
        laf_out = scale_laf(make_upright(new_laf), scale_orig / ellipse_scale)
        return laf_out