File size: 7,664 Bytes
0a82b18
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import os
import torchvision.transforms as tfm
import py3_wget
from matching import BaseMatcher, THIRD_PARTY_DIR, WEIGHTS_DIR
from matching.utils import resize_to_divisible, add_to_path


add_to_path(THIRD_PARTY_DIR.joinpath("DeDoDe"))
from DeDoDe import (
    dedode_detector_L,
    dedode_descriptor_B,
)

add_to_path(THIRD_PARTY_DIR.joinpath("Steerers"))
from rotation_steerers.steerers import DiscreteSteerer, ContinuousSteerer
from rotation_steerers.matchers.max_similarity import (
    MaxSimilarityMatcher,
    ContinuousMaxSimilarityMatcher,
)


class SteererMatcher(BaseMatcher):
    detector_path_L = WEIGHTS_DIR.joinpath("dedode_detector_L.pth")

    descriptor_path_G = WEIGHTS_DIR.joinpath("dedode_descriptor_G.pth")
    descriptor_path_B_C4 = WEIGHTS_DIR.joinpath("B_C4_Perm_descriptor_setting_C.pth")
    descriptor_path_B_SO2 = WEIGHTS_DIR.joinpath("B_SO2_Spread_descriptor_setting_B.pth")

    steerer_path_C = WEIGHTS_DIR.joinpath("B_C4_Perm_steerer_setting_C.pth")
    steerer_path_B = WEIGHTS_DIR.joinpath("B_SO2_Spread_steerer_setting_B.pth")

    dino_patch_size = 14

    def __init__(
        self,
        device="cpu",
        max_num_keypoints=2048,
        dedode_thresh=0.05,
        steerer_type="C8",
        *args,
        **kwargs,
    ):
        super().__init__(device, **kwargs)

        if torch.cuda.is_available() and self.device != "cuda": # only cuda devices work due to autocast in cuda in upstream.
            raise ValueError("Only device 'cuda' supported for Steerers.")

        WEIGHTS_DIR.mkdir(exist_ok=True)
        # download detector
        self.download_weights()

        self.max_keypoints = max_num_keypoints
        self.threshold = dedode_thresh

        self.normalize = tfm.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

        self.detector, self.descriptor, self.steerer, self.matcher = self.build_matcher(steerer_type, device=device)

    def download_weights(self):
        if not os.path.isfile(SteererMatcher.detector_path_L):
            print("Downloading dedode_detector_L.pth")
            py3_wget.download_file(
                "https://github.com/Parskatt/DeDoDe/releases/download/dedode_pretrained_models/dedode_detector_L.pth",
                SteererMatcher.detector_path_L,
            )
        # download descriptors
        if not os.path.isfile(SteererMatcher.descriptor_path_G):
            print("Downloading dedode_descriptor_G.pth")
            py3_wget.download_file(
                "https://github.com/Parskatt/DeDoDe/releases/download/dedode_pretrained_models/dedode_descriptor_G.pth",
                SteererMatcher.descriptor_path_G,
            )
        if not os.path.isfile(SteererMatcher.descriptor_path_B_C4):
            print("Downloading dedode_descriptor_B_C4.pth")
            py3_wget.download_file(
                "https://github.com/georg-bn/rotation-steerers/releases/download/release-2/B_C4_Perm_descriptor_setting_C.pth",
                SteererMatcher.descriptor_path_B_C4,
            )
        if not os.path.isfile(SteererMatcher.descriptor_path_B_SO2):
            print("Downloading dedode_descriptor_B_S02.pth")
            py3_wget.download_file(
                "https://github.com/georg-bn/rotation-steerers/releases/download/release-2/B_SO2_Spread_descriptor_setting_B.pth",
                SteererMatcher.descriptor_path_B_SO2,
            )
        # download steerers
        if not os.path.isfile(SteererMatcher.steerer_path_C):
            print("Downloading B_C4_Perm_steerer_setting_C.pth")
            py3_wget.download_file(
                "https://github.com/georg-bn/rotation-steerers/releases/download/release-2/B_C4_Perm_steerer_setting_C.pth",
                SteererMatcher.steerer_path_C,
            )
        if not os.path.isfile(SteererMatcher.steerer_path_B):
            print("Downloading B_SO2_Spread_steerer_setting_B.pth")
            py3_wget.download_file(
                "https://github.com/georg-bn/rotation-steerers/releases/download/release-2/B_SO2_Spread_steerer_setting_B.pth",
                SteererMatcher.steerer_path_B,
            )

    def build_matcher(self, steerer_type="C8", device="cpu"):
        if steerer_type == "C4":
            detector = dedode_detector_L(weights=torch.load(self.detector_path_L, map_location=device))
            descriptor = dedode_descriptor_B(weights=torch.load(self.descriptor_path_B_C4, map_location=device))
            steerer = DiscreteSteerer(generator=torch.load(self.steerer_path_C, map_location=device))
            steerer_order = 4
        elif steerer_type == "C8":
            detector = dedode_detector_L(weights=torch.load(self.detector_path_L, map_location=device))
            descriptor = dedode_descriptor_B(weights=torch.load(self.descriptor_path_B_SO2, map_location=device))
            steerer_order = 8
            steerer = DiscreteSteerer(
                generator=torch.matrix_exp(
                    (2 * 3.14159 / steerer_order) * torch.load(self.steerer_path_B, map_location=device)
                )
            )

        elif steerer_type == "S02":
            descriptor = dedode_descriptor_B(weights=torch.load(self.descriptor_path_B_SO2, map_location=device))
            steerer = ContinuousSteerer(generator=torch.load(self.steerer_path_B, map_location=device))

        else:
            print(f"Steerer type {steerer_type} not yet implemented")

        if steerer_type == "SO2":
            matcher = ContinuousMaxSimilarityMatcher(steerer=steerer, angles=[0.2, 1.2879, 3.14])
        else:
            matcher = MaxSimilarityMatcher(steerer=steerer, steerer_order=steerer_order)

        return detector, descriptor, steerer, matcher

    def preprocess(self, img):
        # ensure that the img has the proper w/h to be compatible with patch sizes
        _, h, w = img.shape
        orig_shape = h, w
        img = resize_to_divisible(img, self.dino_patch_size)

        img = self.normalize(img).unsqueeze(0).to(self.device)
        return img, orig_shape

    def _forward(self, img0, img1):
        img0, img0_orig_shape = self.preprocess(img0)
        img1, img1_orig_shape = self.preprocess(img1)

        batch_0 = {"image": img0}
        detections_0 = self.detector.detect(batch_0, num_keypoints=self.max_keypoints)
        keypoints_0, P_0 = detections_0["keypoints"], detections_0["confidence"]

        batch_1 = {"image": img1}
        detections_1 = self.detector.detect(batch_1, num_keypoints=self.max_keypoints)
        keypoints_1, P_1 = detections_1["keypoints"], detections_1["confidence"]

        description_0 = self.descriptor.describe_keypoints(batch_0, keypoints_0)["descriptions"]
        description_1 = self.descriptor.describe_keypoints(batch_1, keypoints_1)["descriptions"]

        matches_0, matches_1, _ = self.matcher.match(
            keypoints_0,
            description_0,
            keypoints_1,
            description_1,
            P_A=P_0,
            P_B=P_1,
            normalize=True,
            inv_temp=20,
            threshold=self.threshold,  # Increasing threshold -> fewer matches, fewer outliers
        )

        H0, W0, H1, W1 = *img0.shape[-2:], *img1.shape[-2:]
        mkpts0, mkpts1 = self.matcher.to_pixel_coords(matches_0, matches_1, H0, W0, H1, W1)

        # dedode sometimes requires reshaping an image to fit vit patch size evenly, so we need to
        # rescale kpts to the original img
        mkpts0 = self.rescale_coords(mkpts0, *img0_orig_shape, H0, W0)
        mkpts1 = self.rescale_coords(mkpts1, *img1_orig_shape, H1, W1)

        return mkpts0, mkpts1, keypoints_0, keypoints_1, description_0, description_1