Spaces:
Running
Running
Realcat
commited on
Commit
•
9705edb
1
Parent(s):
78dd8f2
add: doc strings and type
Browse files- app.py +48 -23
- common/utils.py +188 -49
- common/viz.py +151 -37
app.py
CHANGED
@@ -1,5 +1,7 @@
|
|
1 |
import argparse
|
2 |
from pathlib import Path
|
|
|
|
|
3 |
import gradio as gr
|
4 |
from common.utils import (
|
5 |
matcher_zoo,
|
@@ -56,36 +58,59 @@ def ui_change_imagebox(choice):
|
|
56 |
}
|
57 |
|
58 |
|
59 |
-
def ui_reset_state(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
"""
|
61 |
Reset the state of the UI.
|
62 |
|
63 |
Returns:
|
64 |
tuple: A tuple containing the initial values for the UI state.
|
65 |
"""
|
66 |
-
key = list(matcher_zoo.keys())[0] # Get the first key from matcher_zoo
|
67 |
return (
|
68 |
-
None, # image0
|
69 |
-
None, # image1
|
70 |
-
DEFAULT_MATCHING_THRESHOLD, # matching_threshold
|
71 |
-
DEFAULT_SETTING_MAX_FEATURES, # max_features
|
72 |
-
DEFAULT_DEFAULT_KEYPOINT_THRESHOLD, # keypoint_threshold
|
73 |
-
key, # matcher
|
74 |
-
ui_change_imagebox("upload"), # input image0
|
75 |
-
ui_change_imagebox("upload"), # input image1
|
76 |
-
"upload", # match_image_src
|
77 |
-
None, # keypoints
|
78 |
-
None, # raw matches
|
79 |
-
None, # ransac matches
|
80 |
-
{}, # matches result info
|
81 |
-
{}, # matcher config
|
82 |
-
None, # warped
|
83 |
-
{}, # geometry result
|
84 |
-
DEFAULT_RANSAC_METHOD, # ransac_method
|
85 |
-
DEFAULT_RANSAC_REPROJ_THRESHOLD, # ransac_reproj_threshold
|
86 |
-
DEFAULT_RANSAC_CONFIDENCE, # ransac_confidence
|
87 |
-
DEFAULT_RANSAC_MAX_ITER, # ransac_max_iter
|
88 |
-
DEFAULT_SETTING_GEOMETRY, # geometry
|
89 |
)
|
90 |
|
91 |
|
|
|
1 |
import argparse
|
2 |
from pathlib import Path
|
3 |
+
import numpy as np
|
4 |
+
from typing import Dict, Any, Optional, Tuple, List, Union
|
5 |
import gradio as gr
|
6 |
from common.utils import (
|
7 |
matcher_zoo,
|
|
|
58 |
}
|
59 |
|
60 |
|
61 |
+
def ui_reset_state(
|
62 |
+
*args: Any,
|
63 |
+
) -> Tuple[
|
64 |
+
Optional[np.ndarray],
|
65 |
+
Optional[np.ndarray],
|
66 |
+
float,
|
67 |
+
int,
|
68 |
+
float,
|
69 |
+
str,
|
70 |
+
Dict[str, Any],
|
71 |
+
Dict[str, Any],
|
72 |
+
str,
|
73 |
+
Optional[np.ndarray],
|
74 |
+
Optional[np.ndarray],
|
75 |
+
Optional[np.ndarray],
|
76 |
+
Dict[str, Any],
|
77 |
+
Dict[str, Any],
|
78 |
+
Optional[np.ndarray],
|
79 |
+
Dict[str, Any],
|
80 |
+
str,
|
81 |
+
int,
|
82 |
+
float,
|
83 |
+
int,
|
84 |
+
]:
|
85 |
"""
|
86 |
Reset the state of the UI.
|
87 |
|
88 |
Returns:
|
89 |
tuple: A tuple containing the initial values for the UI state.
|
90 |
"""
|
91 |
+
key: str = list(matcher_zoo.keys())[0] # Get the first key from matcher_zoo
|
92 |
return (
|
93 |
+
None, # image0: Optional[np.ndarray]
|
94 |
+
None, # image1: Optional[np.ndarray]
|
95 |
+
DEFAULT_MATCHING_THRESHOLD, # matching_threshold: float
|
96 |
+
DEFAULT_SETTING_MAX_FEATURES, # max_features: int
|
97 |
+
DEFAULT_DEFAULT_KEYPOINT_THRESHOLD, # keypoint_threshold: float
|
98 |
+
key, # matcher: str
|
99 |
+
ui_change_imagebox("upload"), # input image0: Dict[str, Any]
|
100 |
+
ui_change_imagebox("upload"), # input image1: Dict[str, Any]
|
101 |
+
"upload", # match_image_src: str
|
102 |
+
None, # keypoints: Optional[np.ndarray]
|
103 |
+
None, # raw matches: Optional[np.ndarray]
|
104 |
+
None, # ransac matches: Optional[np.ndarray]
|
105 |
+
{}, # matches result info: Dict[str, Any]
|
106 |
+
{}, # matcher config: Dict[str, Any]
|
107 |
+
None, # warped image: Optional[np.ndarray]
|
108 |
+
{}, # geometry result: Dict[str, Any]
|
109 |
+
DEFAULT_RANSAC_METHOD, # ransac_method: str
|
110 |
+
DEFAULT_RANSAC_REPROJ_THRESHOLD, # ransac_reproj_threshold: float
|
111 |
+
DEFAULT_RANSAC_CONFIDENCE, # ransac_confidence: float
|
112 |
+
DEFAULT_RANSAC_MAX_ITER, # ransac_max_iter: int
|
113 |
+
DEFAULT_SETTING_GEOMETRY, # geometry: str
|
114 |
)
|
115 |
|
116 |
|
common/utils.py
CHANGED
@@ -5,6 +5,7 @@ import torch
|
|
5 |
import cv2
|
6 |
import gradio as gr
|
7 |
from pathlib import Path
|
|
|
8 |
from itertools import combinations
|
9 |
from hloc import matchers, extractors, logger
|
10 |
from hloc.utils.base_model import dynamic_load
|
@@ -25,19 +26,39 @@ DEFAULT_RANSAC_MAX_ITER = 10000
|
|
25 |
DEFAULT_MIN_NUM_MATCHES = 4
|
26 |
DEFAULT_MATCHING_THRESHOLD = 0.2
|
27 |
DEFAULT_SETTING_GEOMETRY = "Homography"
|
28 |
-
GRADIO_VERSION = gr.__version__.split(
|
29 |
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
Model = dynamic_load(matchers, match_conf["model"]["name"])
|
32 |
model = Model(match_conf["model"]).eval().to(device)
|
33 |
return model
|
34 |
|
35 |
|
36 |
-
def get_feature_model(conf):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
Model = dynamic_load(extractors, conf["model"]["name"])
|
38 |
model = Model(conf["model"]).eval().to(device)
|
39 |
return model
|
40 |
|
|
|
41 |
def gen_examples():
|
42 |
random.seed(1)
|
43 |
example_matchers = [
|
@@ -92,15 +113,30 @@ def gen_examples():
|
|
92 |
|
93 |
|
94 |
def filter_matches(
|
95 |
-
pred,
|
96 |
-
ransac_method=DEFAULT_RANSAC_METHOD,
|
97 |
-
ransac_reproj_threshold=DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
98 |
-
ransac_confidence=DEFAULT_RANSAC_CONFIDENCE,
|
99 |
-
ransac_max_iter=DEFAULT_RANSAC_MAX_ITER,
|
100 |
-
):
|
101 |
-
|
102 |
-
|
103 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
104 |
if "keypoints0_orig" in pred.keys() and "keypoints1_orig" in pred.keys():
|
105 |
mkpts0 = pred["keypoints0_orig"]
|
106 |
mkpts1 = pred["keypoints1_orig"]
|
@@ -142,20 +178,33 @@ def filter_matches(
|
|
142 |
|
143 |
|
144 |
def compute_geom(
|
145 |
-
pred,
|
146 |
-
ransac_method=DEFAULT_RANSAC_METHOD,
|
147 |
-
ransac_reproj_threshold=DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
148 |
-
ransac_confidence=DEFAULT_RANSAC_CONFIDENCE,
|
149 |
-
ransac_max_iter=DEFAULT_RANSAC_MAX_ITER,
|
150 |
-
) ->
|
151 |
-
|
152 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
153 |
|
154 |
if "keypoints0_orig" in pred.keys() and "keypoints1_orig" in pred.keys():
|
155 |
mkpts0 = pred["keypoints0_orig"]
|
156 |
mkpts1 = pred["keypoints1_orig"]
|
157 |
-
|
158 |
-
if (
|
159 |
"line_keypoints0_orig" in pred.keys()
|
160 |
and "line_keypoints1_orig" in pred.keys()
|
161 |
):
|
@@ -166,7 +215,7 @@ def compute_geom(
|
|
166 |
if len(mkpts0) < 2 * DEFAULT_MIN_NUM_MATCHES:
|
167 |
return {}
|
168 |
h1, w1, _ = pred["image0_orig"].shape
|
169 |
-
geo_info = {}
|
170 |
F, inliers = cv2.findFundamentalMat(
|
171 |
mkpts0,
|
172 |
mkpts1,
|
@@ -197,22 +246,39 @@ def compute_geom(
|
|
197 |
geo_info["H1"] = H1.tolist()
|
198 |
geo_info["H2"] = H2.tolist()
|
199 |
except cv2.error as e:
|
200 |
-
logger.error(f"e, skip")
|
201 |
return geo_info
|
202 |
else:
|
203 |
return {}
|
204 |
|
205 |
|
206 |
-
def wrap_images(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
207 |
h1, w1, _ = img0.shape
|
208 |
h2, w2, _ = img1.shape
|
209 |
-
result_matrix = None
|
210 |
if geo_info is not None and len(geo_info) != 0:
|
211 |
rectified_image0 = img0
|
212 |
rectified_image1 = None
|
213 |
H = np.array(geo_info["Homography"])
|
214 |
F = np.array(geo_info["Fundamental"])
|
215 |
-
title = []
|
216 |
if geom_type == "Homography":
|
217 |
rectified_image1 = cv2.warpPerspective(
|
218 |
img1, H, (img0.shape[1], img0.shape[0])
|
@@ -242,15 +308,32 @@ def wrap_images(img0, img1, geo_info, geom_type):
|
|
242 |
return None, None
|
243 |
|
244 |
|
245 |
-
def change_estimate_geom(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
246 |
if (
|
247 |
matches_info is None
|
248 |
or len(matches_info) < 1
|
249 |
or "geom_info" not in matches_info.keys()
|
250 |
):
|
251 |
return None, None
|
252 |
-
geom_info = matches_info["geom_info"]
|
253 |
-
wrapped_images = None
|
254 |
if choice != "No":
|
255 |
wrapped_images, _ = wrap_images(
|
256 |
input_image0, input_image1, geom_info, choice
|
@@ -260,16 +343,34 @@ def change_estimate_geom(input_image0, input_image1, matches_info, choice):
|
|
260 |
return None, None
|
261 |
|
262 |
|
263 |
-
def display_matches(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
264 |
img0 = pred["image0_orig"]
|
265 |
img1 = pred["image1_orig"]
|
266 |
|
267 |
num_inliers = 0
|
268 |
-
if
|
|
|
|
|
|
|
|
|
|
|
269 |
mkpts0 = pred["keypoints0_orig"]
|
270 |
mkpts1 = pred["keypoints1_orig"]
|
271 |
num_inliers = len(mkpts0)
|
272 |
-
if "mconf" in pred
|
273 |
mconf = pred["mconf"]
|
274 |
else:
|
275 |
mconf = np.ones(len(mkpts0))
|
@@ -283,7 +384,12 @@ def display_matches(pred: dict, titles=[], dpi=300):
|
|
283 |
titles=titles,
|
284 |
)
|
285 |
fig = fig_mkpts
|
286 |
-
if
|
|
|
|
|
|
|
|
|
|
|
287 |
# lines
|
288 |
mtlines0 = pred["line0_orig"]
|
289 |
mtlines1 = pred["line1_orig"]
|
@@ -297,12 +403,12 @@ def display_matches(pred: dict, titles=[], dpi=300):
|
|
297 |
fig_lines = fig2im(fig_lines)
|
298 |
|
299 |
# keypoints
|
300 |
-
mkpts0 = pred
|
301 |
-
mkpts1 = pred
|
302 |
|
303 |
if mkpts0 is not None and mkpts1 is not None:
|
304 |
num_inliers = len(mkpts0)
|
305 |
-
if "mconf" in pred
|
306 |
mconf = pred["mconf"]
|
307 |
else:
|
308 |
mconf = np.ones(len(mkpts0))
|
@@ -317,18 +423,51 @@ def display_matches(pred: dict, titles=[], dpi=300):
|
|
317 |
|
318 |
|
319 |
def run_matching(
|
320 |
-
image0,
|
321 |
-
image1,
|
322 |
-
match_threshold,
|
323 |
-
extract_max_keypoints,
|
324 |
-
keypoint_threshold,
|
325 |
-
key,
|
326 |
-
ransac_method=DEFAULT_RANSAC_METHOD,
|
327 |
-
ransac_reproj_threshold=DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
328 |
-
ransac_confidence=DEFAULT_RANSAC_CONFIDENCE,
|
329 |
-
ransac_max_iter=DEFAULT_RANSAC_MAX_ITER,
|
330 |
-
choice_estimate_geom=DEFAULT_SETTING_GEOMETRY,
|
331 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
332 |
# image0 and image1 is RGB mode
|
333 |
if image0 is None or image1 is None:
|
334 |
raise gr.Error("Error: No images found! Please upload two images.")
|
|
|
5 |
import cv2
|
6 |
import gradio as gr
|
7 |
from pathlib import Path
|
8 |
+
from typing import Dict, Any, Optional, Tuple, List, Union
|
9 |
from itertools import combinations
|
10 |
from hloc import matchers, extractors, logger
|
11 |
from hloc.utils.base_model import dynamic_load
|
|
|
26 |
DEFAULT_MIN_NUM_MATCHES = 4
|
27 |
DEFAULT_MATCHING_THRESHOLD = 0.2
|
28 |
DEFAULT_SETTING_GEOMETRY = "Homography"
|
29 |
+
GRADIO_VERSION = gr.__version__.split(".")[0]
|
30 |
|
31 |
+
|
32 |
+
def get_model(match_conf: Dict[str, Any]):
|
33 |
+
"""
|
34 |
+
Load a matcher model from the provided configuration.
|
35 |
+
|
36 |
+
Args:
|
37 |
+
match_conf: A dictionary containing the model configuration.
|
38 |
+
|
39 |
+
Returns:
|
40 |
+
A matcher model instance.
|
41 |
+
"""
|
42 |
Model = dynamic_load(matchers, match_conf["model"]["name"])
|
43 |
model = Model(match_conf["model"]).eval().to(device)
|
44 |
return model
|
45 |
|
46 |
|
47 |
+
def get_feature_model(conf: Dict[str, Dict[str, Any]]):
|
48 |
+
"""
|
49 |
+
Load a feature extraction model from the provided configuration.
|
50 |
+
|
51 |
+
Args:
|
52 |
+
conf: A dictionary containing the model configuration.
|
53 |
+
|
54 |
+
Returns:
|
55 |
+
A feature extraction model instance.
|
56 |
+
"""
|
57 |
Model = dynamic_load(extractors, conf["model"]["name"])
|
58 |
model = Model(conf["model"]).eval().to(device)
|
59 |
return model
|
60 |
|
61 |
+
|
62 |
def gen_examples():
|
63 |
random.seed(1)
|
64 |
example_matchers = [
|
|
|
113 |
|
114 |
|
115 |
def filter_matches(
|
116 |
+
pred: Dict[str, Any],
|
117 |
+
ransac_method: str = DEFAULT_RANSAC_METHOD,
|
118 |
+
ransac_reproj_threshold: float = DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
119 |
+
ransac_confidence: float = DEFAULT_RANSAC_CONFIDENCE,
|
120 |
+
ransac_max_iter: int = DEFAULT_RANSAC_MAX_ITER,
|
121 |
+
) -> Dict[str, Any]:
|
122 |
+
"""
|
123 |
+
Filter matches using RANSAC. If keypoints are available, filter by keypoints.
|
124 |
+
If lines are available, filter by lines. If both keypoints and lines are
|
125 |
+
available, filter by keypoints.
|
126 |
+
|
127 |
+
Args:
|
128 |
+
pred (Dict[str, Any]): dict of matches, including original keypoints.
|
129 |
+
ransac_method (str, optional): RANSAC method. Defaults to DEFAULT_RANSAC_METHOD.
|
130 |
+
ransac_reproj_threshold (float, optional): RANSAC reprojection threshold. Defaults to DEFAULT_RANSAC_REPROJ_THRESHOLD.
|
131 |
+
ransac_confidence (float, optional): RANSAC confidence. Defaults to DEFAULT_RANSAC_CONFIDENCE.
|
132 |
+
ransac_max_iter (int, optional): RANSAC maximum iterations. Defaults to DEFAULT_RANSAC_MAX_ITER.
|
133 |
+
|
134 |
+
Returns:
|
135 |
+
Dict[str, Any]: filtered matches.
|
136 |
+
"""
|
137 |
+
mkpts0: Optional[np.ndarray] = None
|
138 |
+
mkpts1: Optional[np.ndarray] = None
|
139 |
+
feature_type: Optional[str] = None
|
140 |
if "keypoints0_orig" in pred.keys() and "keypoints1_orig" in pred.keys():
|
141 |
mkpts0 = pred["keypoints0_orig"]
|
142 |
mkpts1 = pred["keypoints1_orig"]
|
|
|
178 |
|
179 |
|
180 |
def compute_geom(
|
181 |
+
pred: Dict[str, Any],
|
182 |
+
ransac_method: str = DEFAULT_RANSAC_METHOD,
|
183 |
+
ransac_reproj_threshold: float = DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
184 |
+
ransac_confidence: float = DEFAULT_RANSAC_CONFIDENCE,
|
185 |
+
ransac_max_iter: int = DEFAULT_RANSAC_MAX_ITER,
|
186 |
+
) -> Dict[str, List[float]]:
|
187 |
+
"""
|
188 |
+
Compute geometric information of matches, including Fundamental matrix,
|
189 |
+
Homography matrix, and rectification matrices (if available).
|
190 |
+
|
191 |
+
Args:
|
192 |
+
pred (Dict[str, Any]): dict of matches, including original keypoints.
|
193 |
+
ransac_method (str, optional): RANSAC method. Defaults to DEFAULT_RANSAC_METHOD.
|
194 |
+
ransac_reproj_threshold (float, optional): RANSAC reprojection threshold. Defaults to DEFAULT_RANSAC_REPROJ_THRESHOLD.
|
195 |
+
ransac_confidence (float, optional): RANSAC confidence. Defaults to DEFAULT_RANSAC_CONFIDENCE.
|
196 |
+
ransac_max_iter (int, optional): RANSAC maximum iterations. Defaults to DEFAULT_RANSAC_MAX_ITER.
|
197 |
+
|
198 |
+
Returns:
|
199 |
+
Dict[str, List[float]]: geometric information in form of a dict.
|
200 |
+
"""
|
201 |
+
mkpts0: Optional[np.ndarray] = None
|
202 |
+
mkpts1: Optional[np.ndarray] = None
|
203 |
|
204 |
if "keypoints0_orig" in pred.keys() and "keypoints1_orig" in pred.keys():
|
205 |
mkpts0 = pred["keypoints0_orig"]
|
206 |
mkpts1 = pred["keypoints1_orig"]
|
207 |
+
elif (
|
|
|
208 |
"line_keypoints0_orig" in pred.keys()
|
209 |
and "line_keypoints1_orig" in pred.keys()
|
210 |
):
|
|
|
215 |
if len(mkpts0) < 2 * DEFAULT_MIN_NUM_MATCHES:
|
216 |
return {}
|
217 |
h1, w1, _ = pred["image0_orig"].shape
|
218 |
+
geo_info: Dict[str, List[float]] = {}
|
219 |
F, inliers = cv2.findFundamentalMat(
|
220 |
mkpts0,
|
221 |
mkpts1,
|
|
|
246 |
geo_info["H1"] = H1.tolist()
|
247 |
geo_info["H2"] = H2.tolist()
|
248 |
except cv2.error as e:
|
249 |
+
logger.error(f"{e}, skip")
|
250 |
return geo_info
|
251 |
else:
|
252 |
return {}
|
253 |
|
254 |
|
255 |
+
def wrap_images(
|
256 |
+
img0: np.ndarray,
|
257 |
+
img1: np.ndarray,
|
258 |
+
geo_info: Optional[Dict[str, List[float]]],
|
259 |
+
geom_type: str,
|
260 |
+
) -> Tuple[Optional[str], Optional[Dict[str, List[float]]]]:
|
261 |
+
"""
|
262 |
+
Wraps the images based on the geometric transformation used to align them.
|
263 |
+
|
264 |
+
Args:
|
265 |
+
img0: numpy array representing the first image.
|
266 |
+
img1: numpy array representing the second image.
|
267 |
+
geo_info: dictionary containing the geometric transformation information.
|
268 |
+
geom_type: type of geometric transformation used to align the images.
|
269 |
+
|
270 |
+
Returns:
|
271 |
+
A tuple containing a base64 encoded image string and a dictionary with the transformation matrix.
|
272 |
+
"""
|
273 |
h1, w1, _ = img0.shape
|
274 |
h2, w2, _ = img1.shape
|
275 |
+
result_matrix: Optional[np.ndarray] = None
|
276 |
if geo_info is not None and len(geo_info) != 0:
|
277 |
rectified_image0 = img0
|
278 |
rectified_image1 = None
|
279 |
H = np.array(geo_info["Homography"])
|
280 |
F = np.array(geo_info["Fundamental"])
|
281 |
+
title: List[str] = []
|
282 |
if geom_type == "Homography":
|
283 |
rectified_image1 = cv2.warpPerspective(
|
284 |
img1, H, (img0.shape[1], img0.shape[0])
|
|
|
308 |
return None, None
|
309 |
|
310 |
|
311 |
+
def change_estimate_geom(
|
312 |
+
input_image0: np.ndarray,
|
313 |
+
input_image1: np.ndarray,
|
314 |
+
matches_info: Dict[str, Any],
|
315 |
+
choice: str,
|
316 |
+
) -> Tuple[Optional[np.ndarray], Optional[Dict[str, Any]]]:
|
317 |
+
"""
|
318 |
+
Changes the estimate of the geometric transformation used to align the images.
|
319 |
+
|
320 |
+
Args:
|
321 |
+
input_image0: First input image.
|
322 |
+
input_image1: Second input image.
|
323 |
+
matches_info: Dictionary containing information about the matches.
|
324 |
+
choice: Type of geometric transformation to use ('Homography' or 'Fundamental') or 'No' to disable.
|
325 |
+
|
326 |
+
Returns:
|
327 |
+
A tuple containing the updated images and the updated matches info.
|
328 |
+
"""
|
329 |
if (
|
330 |
matches_info is None
|
331 |
or len(matches_info) < 1
|
332 |
or "geom_info" not in matches_info.keys()
|
333 |
):
|
334 |
return None, None
|
335 |
+
geom_info: Dict[str, Any] = matches_info["geom_info"]
|
336 |
+
wrapped_images: Optional[np.ndarray] = None
|
337 |
if choice != "No":
|
338 |
wrapped_images, _ = wrap_images(
|
339 |
input_image0, input_image1, geom_info, choice
|
|
|
343 |
return None, None
|
344 |
|
345 |
|
346 |
+
def display_matches(
|
347 |
+
pred: Dict[str, np.ndarray], titles: List[str] = [], dpi: int = 300
|
348 |
+
) -> Tuple[np.ndarray, int]:
|
349 |
+
"""
|
350 |
+
Displays the matches between two images.
|
351 |
+
|
352 |
+
Args:
|
353 |
+
pred: Dictionary containing the original images and the matches.
|
354 |
+
titles: Optional titles for the plot.
|
355 |
+
dpi: Resolution of the plot.
|
356 |
+
|
357 |
+
Returns:
|
358 |
+
The resulting concatenated plot and the number of inliers.
|
359 |
+
"""
|
360 |
img0 = pred["image0_orig"]
|
361 |
img1 = pred["image1_orig"]
|
362 |
|
363 |
num_inliers = 0
|
364 |
+
if (
|
365 |
+
"keypoints0_orig" in pred
|
366 |
+
and "keypoints1_orig" in pred
|
367 |
+
and pred["keypoints0_orig"] is not None
|
368 |
+
and pred["keypoints1_orig"] is not None
|
369 |
+
):
|
370 |
mkpts0 = pred["keypoints0_orig"]
|
371 |
mkpts1 = pred["keypoints1_orig"]
|
372 |
num_inliers = len(mkpts0)
|
373 |
+
if "mconf" in pred:
|
374 |
mconf = pred["mconf"]
|
375 |
else:
|
376 |
mconf = np.ones(len(mkpts0))
|
|
|
384 |
titles=titles,
|
385 |
)
|
386 |
fig = fig_mkpts
|
387 |
+
if (
|
388 |
+
"line0_orig" in pred
|
389 |
+
and "line1_orig" in pred
|
390 |
+
and pred["line0_orig"] is not None
|
391 |
+
and pred["line1_orig"] is not None
|
392 |
+
):
|
393 |
# lines
|
394 |
mtlines0 = pred["line0_orig"]
|
395 |
mtlines1 = pred["line1_orig"]
|
|
|
403 |
fig_lines = fig2im(fig_lines)
|
404 |
|
405 |
# keypoints
|
406 |
+
mkpts0 = pred.get("line_keypoints0_orig")
|
407 |
+
mkpts1 = pred.get("line_keypoints1_orig")
|
408 |
|
409 |
if mkpts0 is not None and mkpts1 is not None:
|
410 |
num_inliers = len(mkpts0)
|
411 |
+
if "mconf" in pred:
|
412 |
mconf = pred["mconf"]
|
413 |
else:
|
414 |
mconf = np.ones(len(mkpts0))
|
|
|
423 |
|
424 |
|
425 |
def run_matching(
|
426 |
+
image0: np.ndarray,
|
427 |
+
image1: np.ndarray,
|
428 |
+
match_threshold: float,
|
429 |
+
extract_max_keypoints: int,
|
430 |
+
keypoint_threshold: float,
|
431 |
+
key: str,
|
432 |
+
ransac_method: str = DEFAULT_RANSAC_METHOD,
|
433 |
+
ransac_reproj_threshold: int = DEFAULT_RANSAC_REPROJ_THRESHOLD,
|
434 |
+
ransac_confidence: float = DEFAULT_RANSAC_CONFIDENCE,
|
435 |
+
ransac_max_iter: int = DEFAULT_RANSAC_MAX_ITER,
|
436 |
+
choice_estimate_geom: str = DEFAULT_SETTING_GEOMETRY,
|
437 |
+
) -> Tuple[
|
438 |
+
np.ndarray,
|
439 |
+
np.ndarray,
|
440 |
+
np.ndarray,
|
441 |
+
Dict[str, int],
|
442 |
+
Dict[str, Dict[str, Any]],
|
443 |
+
Dict[str, Dict[str, float]],
|
444 |
+
np.ndarray,
|
445 |
+
]:
|
446 |
+
"""Match two images using the given parameters.
|
447 |
+
|
448 |
+
Args:
|
449 |
+
image0 (np.ndarray): RGB image 0.
|
450 |
+
image1 (np.ndarray): RGB image 1.
|
451 |
+
match_threshold (float): match threshold.
|
452 |
+
extract_max_keypoints (int): number of keypoints to extract.
|
453 |
+
keypoint_threshold (float): keypoint threshold.
|
454 |
+
key (str): key of the model to use.
|
455 |
+
ransac_method (str, optional): RANSAC method to use.
|
456 |
+
ransac_reproj_threshold (int, optional): RANSAC reprojection threshold.
|
457 |
+
ransac_confidence (float, optional): RANSAC confidence level.
|
458 |
+
ransac_max_iter (int, optional): RANSAC maximum number of iterations.
|
459 |
+
choice_estimate_geom (str, optional): setting of geometry estimation.
|
460 |
+
|
461 |
+
Returns:
|
462 |
+
tuple:
|
463 |
+
- output_keypoints (np.ndarray): image with keypoints.
|
464 |
+
- output_matches_raw (np.ndarray): image with raw matches.
|
465 |
+
- output_matches_ransac (np.ndarray): image with RANSAC matches.
|
466 |
+
- num_matches (Dict[str, int]): number of raw and RANSAC matches.
|
467 |
+
- configs (Dict[str, Dict[str, Any]]): match and feature extraction configs.
|
468 |
+
- geom_info (Dict[str, Dict[str, float]]): geometry information.
|
469 |
+
- output_wrapped (np.ndarray): wrapped images.
|
470 |
+
"""
|
471 |
# image0 and image1 is RGB mode
|
472 |
if image0 is None or image1 is None:
|
473 |
raise gr.Error("Error: No images found! Please upload two images.")
|
common/viz.py
CHANGED
@@ -1,20 +1,35 @@
|
|
1 |
import numpy as np
|
2 |
-
import matplotlib.pyplot as plt
|
3 |
-
import matplotlib
|
4 |
import seaborn as sns
|
|
|
|
|
|
|
|
|
5 |
|
6 |
|
7 |
-
def plot_images(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
"""Plot a set of images horizontally.
|
9 |
Args:
|
10 |
imgs: a list of NumPy or PyTorch images, RGB (H, W, 3) or mono (H, W).
|
11 |
titles: a list of strings, as titles for each image.
|
12 |
-
cmaps: colormaps for monochrome images.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
"""
|
14 |
n = len(imgs)
|
15 |
-
if not isinstance(cmaps,
|
16 |
cmaps = [cmaps] * n
|
17 |
-
# figsize = (size*n, size*3/4) if size is not None else None
|
18 |
figsize = (size * n, size * 6 / 5) if size is not None else None
|
19 |
fig, ax = plt.subplots(1, n, figsize=figsize, dpi=dpi)
|
20 |
|
@@ -33,24 +48,33 @@ def plot_images(imgs, titles=None, cmaps="gray", dpi=100, size=5, pad=0.5):
|
|
33 |
return fig
|
34 |
|
35 |
|
36 |
-
def plot_color_line_matches(
|
|
|
|
|
|
|
|
|
|
|
37 |
"""Plot line matches for existing images with multiple colors.
|
|
|
38 |
Args:
|
39 |
-
lines:
|
40 |
-
correct_matches: bool array of size (N,) indicating correct
|
41 |
-
|
42 |
-
|
|
|
|
|
|
|
|
|
43 |
"""
|
44 |
-
n_lines =
|
45 |
colors = sns.color_palette("husl", n_colors=n_lines)
|
46 |
np.random.shuffle(colors)
|
47 |
alphas = np.ones(n_lines)
|
48 |
-
# If correct_matches is not None, display wrong matches with a low alpha
|
49 |
if correct_matches is not None:
|
50 |
alphas[~np.array(correct_matches)] = 0.2
|
51 |
|
52 |
fig = plt.gcf()
|
53 |
-
ax = fig.axes
|
54 |
assert len(ax) > max(indices)
|
55 |
axes = [ax[i] for i in indices]
|
56 |
fig.canvas.draw()
|
@@ -78,21 +102,39 @@ def plot_color_line_matches(lines, correct_matches=None, lw=2, indices=(0, 1)):
|
|
78 |
|
79 |
|
80 |
def make_matching_figure(
|
81 |
-
img0,
|
82 |
-
img1,
|
83 |
-
mkpts0,
|
84 |
-
mkpts1,
|
85 |
-
color,
|
86 |
-
titles=None,
|
87 |
-
kpts0=None,
|
88 |
-
kpts1=None,
|
89 |
-
text=[],
|
90 |
-
dpi=75,
|
91 |
-
path=None,
|
92 |
-
pad=0,
|
93 |
-
):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
94 |
# draw image pair
|
95 |
-
# assert mkpts0.shape[0] == mkpts1.shape[0], f'mkpts0: {mkpts0.shape[0]} v.s. mkpts1: {mkpts1.shape[0]}'
|
96 |
fig, axes = plt.subplots(1, 2, figsize=(10, 6), dpi=dpi)
|
97 |
axes[0].imshow(img0) # , cmap='gray')
|
98 |
axes[1].imshow(img1) # , cmap='gray')
|
@@ -156,7 +198,20 @@ def make_matching_figure(
|
|
156 |
return fig
|
157 |
|
158 |
|
159 |
-
def error_colormap(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
160 |
assert alpha <= 1.0 and alpha > 0, f"Invaid alpha value: {alpha}"
|
161 |
x = 1 - np.clip(err / (thr * 2), 0, 1)
|
162 |
return np.clip(
|
@@ -173,22 +228,57 @@ color_map = np.arange(100)
|
|
173 |
np.random.shuffle(color_map)
|
174 |
|
175 |
|
176 |
-
def fig2im(fig):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
fig.canvas.draw()
|
178 |
-
|
179 |
buf_ndarray = np.frombuffer(fig.canvas.tostring_rgb(), dtype="u1")
|
180 |
-
|
181 |
-
return im
|
182 |
|
183 |
|
184 |
def draw_matches(
|
185 |
-
mkpts0
|
186 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
187 |
thr = 5e-4
|
188 |
thr = 0.5
|
189 |
color = error_colormap(conf, thr, alpha=0.1)
|
190 |
text = [
|
191 |
-
|
192 |
f"#Matches: {len(mkpts0)}",
|
193 |
]
|
194 |
if path:
|
@@ -222,7 +312,31 @@ def draw_matches(
|
|
222 |
)
|
223 |
|
224 |
|
225 |
-
def draw_image_pairs(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
226 |
# draw image pair
|
227 |
fig, axes = plt.subplots(1, 2, figsize=(10, 6), dpi=dpi)
|
228 |
axes[0].imshow(img0) # , cmap='gray')
|
|
|
1 |
import numpy as np
|
|
|
|
|
2 |
import seaborn as sns
|
3 |
+
import matplotlib
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
+
from pathlib import Path
|
6 |
+
from typing import Dict, Any, Optional, Tuple, List, Union
|
7 |
|
8 |
|
9 |
+
def plot_images(
|
10 |
+
imgs: List[np.ndarray],
|
11 |
+
titles: Optional[List[str]] = None,
|
12 |
+
cmaps: Union[str, List[str]] = "gray",
|
13 |
+
dpi: int = 100,
|
14 |
+
size: Optional[int] = 5,
|
15 |
+
pad: float = 0.5,
|
16 |
+
) -> plt.Figure:
|
17 |
"""Plot a set of images horizontally.
|
18 |
Args:
|
19 |
imgs: a list of NumPy or PyTorch images, RGB (H, W, 3) or mono (H, W).
|
20 |
titles: a list of strings, as titles for each image.
|
21 |
+
cmaps: colormaps for monochrome images. If a single string is given,
|
22 |
+
it is used for all images.
|
23 |
+
dpi: DPI of the figure.
|
24 |
+
size: figure size in inches (width). If not provided, the figure
|
25 |
+
size is determined automatically.
|
26 |
+
pad: padding between subplots, in inches.
|
27 |
+
Returns:
|
28 |
+
The created figure.
|
29 |
"""
|
30 |
n = len(imgs)
|
31 |
+
if not isinstance(cmaps, list):
|
32 |
cmaps = [cmaps] * n
|
|
|
33 |
figsize = (size * n, size * 6 / 5) if size is not None else None
|
34 |
fig, ax = plt.subplots(1, n, figsize=figsize, dpi=dpi)
|
35 |
|
|
|
48 |
return fig
|
49 |
|
50 |
|
51 |
+
def plot_color_line_matches(
|
52 |
+
lines: List[np.ndarray],
|
53 |
+
correct_matches: Optional[np.ndarray] = None,
|
54 |
+
lw: float = 2.0,
|
55 |
+
indices: Tuple[int, int] = (0, 1),
|
56 |
+
) -> matplotlib.figure.Figure:
|
57 |
"""Plot line matches for existing images with multiple colors.
|
58 |
+
|
59 |
Args:
|
60 |
+
lines: List of ndarrays of size (N, 2, 2) representing line segments.
|
61 |
+
correct_matches: Optional bool array of size (N,) indicating correct
|
62 |
+
matches. If not None, display wrong matches with a low alpha.
|
63 |
+
lw: Line width as float pixels.
|
64 |
+
indices: Indices of the images to draw the matches on.
|
65 |
+
|
66 |
+
Returns:
|
67 |
+
The modified matplotlib figure.
|
68 |
"""
|
69 |
+
n_lines = lines[0].shape[0]
|
70 |
colors = sns.color_palette("husl", n_colors=n_lines)
|
71 |
np.random.shuffle(colors)
|
72 |
alphas = np.ones(n_lines)
|
|
|
73 |
if correct_matches is not None:
|
74 |
alphas[~np.array(correct_matches)] = 0.2
|
75 |
|
76 |
fig = plt.gcf()
|
77 |
+
ax = typing.cast(List[matplotlib.axes.Axes], fig.axes)
|
78 |
assert len(ax) > max(indices)
|
79 |
axes = [ax[i] for i in indices]
|
80 |
fig.canvas.draw()
|
|
|
102 |
|
103 |
|
104 |
def make_matching_figure(
|
105 |
+
img0: np.ndarray,
|
106 |
+
img1: np.ndarray,
|
107 |
+
mkpts0: np.ndarray,
|
108 |
+
mkpts1: np.ndarray,
|
109 |
+
color: np.ndarray,
|
110 |
+
titles: Optional[List[str]] = None,
|
111 |
+
kpts0: Optional[np.ndarray] = None,
|
112 |
+
kpts1: Optional[np.ndarray] = None,
|
113 |
+
text: List[str] = [],
|
114 |
+
dpi: int = 75,
|
115 |
+
path: Optional[Path] = None,
|
116 |
+
pad: float = 0.0,
|
117 |
+
) -> Optional[plt.Figure]:
|
118 |
+
"""Draw image pair with matches.
|
119 |
+
|
120 |
+
Args:
|
121 |
+
img0: image0 as HxWx3 numpy array.
|
122 |
+
img1: image1 as HxWx3 numpy array.
|
123 |
+
mkpts0: matched points in image0 as Nx2 numpy array.
|
124 |
+
mkpts1: matched points in image1 as Nx2 numpy array.
|
125 |
+
color: colors for the matches as Nx4 numpy array.
|
126 |
+
titles: titles for the two subplots.
|
127 |
+
kpts0: keypoints in image0 as Kx2 numpy array.
|
128 |
+
kpts1: keypoints in image1 as Kx2 numpy array.
|
129 |
+
text: list of strings to display in the top-left corner of the image.
|
130 |
+
dpi: dots per inch of the saved figure.
|
131 |
+
path: if not None, save the figure to this path.
|
132 |
+
pad: padding around the image as a fraction of the image size.
|
133 |
+
|
134 |
+
Returns:
|
135 |
+
The matplotlib Figure object if path is None.
|
136 |
+
"""
|
137 |
# draw image pair
|
|
|
138 |
fig, axes = plt.subplots(1, 2, figsize=(10, 6), dpi=dpi)
|
139 |
axes[0].imshow(img0) # , cmap='gray')
|
140 |
axes[1].imshow(img1) # , cmap='gray')
|
|
|
198 |
return fig
|
199 |
|
200 |
|
201 |
+
def error_colormap(
|
202 |
+
err: np.ndarray, thr: float, alpha: float = 1.0
|
203 |
+
) -> np.ndarray:
|
204 |
+
"""
|
205 |
+
Create a colormap based on the error values.
|
206 |
+
|
207 |
+
Args:
|
208 |
+
err: Error values as a numpy array of shape (N,).
|
209 |
+
thr: Threshold value for the error.
|
210 |
+
alpha: Alpha value for the colormap, between 0 and 1.
|
211 |
+
|
212 |
+
Returns:
|
213 |
+
Colormap as a numpy array of shape (N, 4) with values in [0, 1].
|
214 |
+
"""
|
215 |
assert alpha <= 1.0 and alpha > 0, f"Invaid alpha value: {alpha}"
|
216 |
x = 1 - np.clip(err / (thr * 2), 0, 1)
|
217 |
return np.clip(
|
|
|
228 |
np.random.shuffle(color_map)
|
229 |
|
230 |
|
231 |
+
def fig2im(fig: matplotlib.figure.Figure) -> np.ndarray:
|
232 |
+
"""
|
233 |
+
Convert a matplotlib figure to a numpy array with RGB values.
|
234 |
+
|
235 |
+
Args:
|
236 |
+
fig: A matplotlib figure.
|
237 |
+
|
238 |
+
Returns:
|
239 |
+
A numpy array with shape (height, width, 3) and dtype uint8 containing
|
240 |
+
the RGB values of the figure.
|
241 |
+
"""
|
242 |
fig.canvas.draw()
|
243 |
+
(width, height) = fig.canvas.get_width_height()
|
244 |
buf_ndarray = np.frombuffer(fig.canvas.tostring_rgb(), dtype="u1")
|
245 |
+
return buf_ndarray.reshape(height, width, 3)
|
|
|
246 |
|
247 |
|
248 |
def draw_matches(
|
249 |
+
mkpts0: List[np.ndarray],
|
250 |
+
mkpts1: List[np.ndarray],
|
251 |
+
img0: np.ndarray,
|
252 |
+
img1: np.ndarray,
|
253 |
+
conf: np.ndarray,
|
254 |
+
titles: Optional[List[str]] = None,
|
255 |
+
dpi: int = 150,
|
256 |
+
path: Optional[str] = None,
|
257 |
+
pad: float = 0.5,
|
258 |
+
) -> np.ndarray:
|
259 |
+
"""
|
260 |
+
Draw matches between two images.
|
261 |
+
|
262 |
+
Args:
|
263 |
+
mkpts0: List of matches from the first image, with shape (N, 2)
|
264 |
+
mkpts1: List of matches from the second image, with shape (N, 2)
|
265 |
+
img0: First image, with shape (H, W, 3)
|
266 |
+
img1: Second image, with shape (H, W, 3)
|
267 |
+
conf: Confidence values for the matches, with shape (N,)
|
268 |
+
titles: Optional list of title strings for the plot
|
269 |
+
dpi: DPI for the saved image
|
270 |
+
path: Optional path to save the image to. If None, the image is not saved.
|
271 |
+
pad: Padding between subplots
|
272 |
+
|
273 |
+
Returns:
|
274 |
+
The figure as a numpy array with shape (height, width, 3) and dtype uint8
|
275 |
+
containing the RGB values of the figure.
|
276 |
+
"""
|
277 |
thr = 5e-4
|
278 |
thr = 0.5
|
279 |
color = error_colormap(conf, thr, alpha=0.1)
|
280 |
text = [
|
281 |
+
"image name",
|
282 |
f"#Matches: {len(mkpts0)}",
|
283 |
]
|
284 |
if path:
|
|
|
312 |
)
|
313 |
|
314 |
|
315 |
+
def draw_image_pairs(
|
316 |
+
img0: np.ndarray,
|
317 |
+
img1: np.ndarray,
|
318 |
+
text: List[str] = [],
|
319 |
+
dpi: int = 75,
|
320 |
+
path: Optional[str] = None,
|
321 |
+
pad: float = 0.5,
|
322 |
+
) -> np.ndarray:
|
323 |
+
"""Draw image pair horizontally.
|
324 |
+
|
325 |
+
Args:
|
326 |
+
img0: First image, with shape (H, W, 3)
|
327 |
+
img1: Second image, with shape (H, W, 3)
|
328 |
+
text: List of strings to print. Each string is a new line.
|
329 |
+
dpi: DPI of the figure.
|
330 |
+
path: Path to save the image to. If None, the image is not saved and
|
331 |
+
the function returns the figure as a numpy array with shape
|
332 |
+
(height, width, 3) and dtype uint8 containing the RGB values of the
|
333 |
+
figure.
|
334 |
+
pad: Padding between subplots
|
335 |
+
|
336 |
+
Returns:
|
337 |
+
The figure as a numpy array with shape (height, width, 3) and dtype uint8
|
338 |
+
containing the RGB values of the figure, or None if path is not None.
|
339 |
+
"""
|
340 |
# draw image pair
|
341 |
fig, axes = plt.subplots(1, 2, figsize=(10, 6), dpi=dpi)
|
342 |
axes[0].imshow(img0) # , cmap='gray')
|