Commit
·
81e7fbf
1
Parent(s):
e398178
Updated Utility Files
Browse files- README.md +5 -0
- classes_color.json +3 -0
- consistency.py +123 -56
- loader.py +29 -12
- requirements.txt +14 -0
- segmentation.py +8 -28
- utils.py +19 -4
README.md
CHANGED
@@ -28,6 +28,7 @@ gtdh-hd
|
|
28 |
| loader.py # Simple Dataset Loader and Storage Functions
|
29 |
│ segmentation.py # Multiclass Segmentation Generation
|
30 |
│ utils.py # Helper Functions
|
|
|
31 |
└───drafter_D
|
32 |
│ └───annotations # Bounding Box Annotations
|
33 |
│ │ │ CX_DY_PZ.xml
|
@@ -305,3 +306,7 @@ Rotation annotations are currently work in progress. They should be provided for
|
|
305 |
```
|
306 |
labelme --labels "connector" --config "{shift_auto_shape_color: 1}" --nodata
|
307 |
```
|
|
|
|
|
|
|
|
|
|
28 |
| loader.py # Simple Dataset Loader and Storage Functions
|
29 |
│ segmentation.py # Multiclass Segmentation Generation
|
30 |
│ utils.py # Helper Functions
|
31 |
+
│ requirements.txt # Requirements for Scripts
|
32 |
└───drafter_D
|
33 |
│ └───annotations # Bounding Box Annotations
|
34 |
│ │ │ CX_DY_PZ.xml
|
|
|
306 |
```
|
307 |
labelme --labels "connector" --config "{shift_auto_shape_color: 1}" --nodata
|
308 |
```
|
309 |
+
|
310 |
+
## Licence
|
311 |
+
The entire content of this repository, including all image files, annotation files as well as has sourcecode, metadata and documentation has been published under the [Creative Commons Attribution Share Alike Licence 3.0](https://creativecommons.org/licenses/by-sa/3.0/).
|
312 |
+
|
classes_color.json
CHANGED
@@ -69,5 +69,8 @@
|
|
69 |
"antenna": [112,128,144],
|
70 |
"crystal": [230,230,250],
|
71 |
|
|
|
|
|
|
|
72 |
"unknown": [240,255,240]
|
73 |
}
|
|
|
69 |
"antenna": [112,128,144],
|
70 |
"crystal": [230,230,250],
|
71 |
|
72 |
+
"magnetic": [0,230,230],
|
73 |
+
"optical": [230,0,230],
|
74 |
+
|
75 |
"unknown": [240,255,240]
|
76 |
}
|
consistency.py
CHANGED
@@ -7,6 +7,7 @@ import re
|
|
7 |
|
8 |
# Project Imports
|
9 |
from loader import load_classes, load_properties, read_dataset, write_dataset, file_name
|
|
|
10 |
|
11 |
# Third-Party Imports
|
12 |
import matplotlib.pyplot as plt
|
@@ -29,41 +30,42 @@ MAPPING_LOOKUP = {
|
|
29 |
}
|
30 |
|
31 |
|
32 |
-
def consistency(db: list, classes: dict, recover: dict = {}) -> tuple:
|
33 |
"""Checks Whether Annotation Classes are in provided Classes Dict and Attempts Recovery"""
|
34 |
|
35 |
total, ok, mapped, faulty, rotation, text = 0, 0, 0, 0, 0, 0
|
36 |
|
37 |
for sample in db:
|
38 |
-
for
|
39 |
total += 1
|
40 |
|
41 |
-
if
|
42 |
ok += 1
|
43 |
|
44 |
-
if
|
45 |
-
|
46 |
mapped += 1
|
47 |
|
48 |
-
if
|
49 |
-
print(f"Can't recover faulty label in {file_name(sample)}: {
|
50 |
faulty += 1
|
51 |
|
52 |
-
if
|
53 |
rotation += 1
|
54 |
|
55 |
-
if
|
56 |
-
|
|
|
57 |
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
|
63 |
-
|
64 |
-
|
65 |
|
66 |
-
|
67 |
|
68 |
return total, ok, mapped, faulty, rotation, text
|
69 |
|
@@ -109,12 +111,21 @@ def circuit_annotations(db: list, classes: dict) -> None:
|
|
109 |
plt.show()
|
110 |
|
111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
def class_distribution(db: list, classes: dict) -> None:
|
113 |
"""Plots the Class Distribution over the Dataset"""
|
114 |
|
115 |
class_nbrs = np.arange(len(classes))
|
116 |
-
class_counts = [sum([len([
|
117 |
-
if
|
118 |
for sample in db]) for cls in classes]
|
119 |
|
120 |
bars = plt.bar(class_nbrs, class_counts)
|
@@ -154,12 +165,15 @@ def image_count(drafter: int = None, segmentation: bool = False) -> int:
|
|
154 |
(not drafter or f"drafter_{drafter}{os.sep}" in root)])
|
155 |
|
156 |
|
157 |
-
def read_check_write(classes: dict, drafter: int = None, segmentation: bool = False):
|
158 |
"""Reads Annotations, Checks Consistency with Provided Classes
|
159 |
Writes Corrected Annotations Back and Returns the Annotations"""
|
160 |
|
161 |
db = read_dataset(drafter=drafter, segmentation=segmentation)
|
162 |
-
ann_total, ann_ok, ann_mapped, ann_faulty, ann_rot, ann_text = consistency(db,
|
|
|
|
|
|
|
163 |
write_dataset(db, segmentation=segmentation)
|
164 |
|
165 |
print("")
|
@@ -179,53 +193,101 @@ def read_check_write(classes: dict, drafter: int = None, segmentation: bool = Fa
|
|
179 |
return db
|
180 |
|
181 |
|
182 |
-
def
|
183 |
-
"""
|
184 |
-
|
185 |
-
print("")
|
186 |
-
print(" Text Statistics")
|
187 |
-
print("---------------------")
|
188 |
-
|
189 |
-
text_bbs = len([bbox for sample in db for bbox in sample["bboxes"] if bbox["class"] == "text"])
|
190 |
-
print(f"Text BB Annotations: {text_bbs}")
|
191 |
-
|
192 |
-
text_labels = [bbox["text"] for sample in db for bbox in sample["bboxes"] if bbox["text"] is not None]
|
193 |
-
print(f"Overall Text Label Count: {len(text_labels)}")
|
194 |
|
195 |
-
|
196 |
-
|
197 |
|
198 |
-
print(f"Total Character Count: {sum([len(text_label) for text_label in text_labels])}")
|
199 |
|
200 |
-
|
201 |
-
|
202 |
-
chars = sorted(list(char_set))
|
203 |
-
print(chars)
|
204 |
|
205 |
char_nbrs = np.arange(len(chars))
|
206 |
char_counts = [sum([len([None for text_char in text_label if text_char == char])
|
207 |
-
for text_label in
|
208 |
for char in chars]
|
209 |
plt.bar(char_nbrs, char_counts)
|
210 |
plt.xticks(char_nbrs, chars)
|
211 |
-
plt.title("Character
|
212 |
plt.xlabel("Character")
|
213 |
plt.ylabel("Overall Count")
|
214 |
plt.show()
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
plt.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
227 |
plt.show()
|
228 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
229 |
text_instances = text_labels_unique if plot_unique_labels else text_labels
|
230 |
text_classes_names = []
|
231 |
text_classes_instances = []
|
@@ -252,18 +314,23 @@ def text_statistics(db: list, plot_unique_labels: bool = False):
|
|
252 |
plt.tight_layout()
|
253 |
plt.show()
|
254 |
|
|
|
|
|
|
|
|
|
255 |
|
256 |
|
257 |
if __name__ == "__main__":
|
258 |
drafter_selected = int(sys.argv[1]) if len(sys.argv) == 2 else None
|
259 |
classes = load_classes()
|
260 |
|
261 |
-
#db_poly = read_check_write(classes, drafter_selected, segmentation=True)
|
262 |
db_bb = read_check_write(classes, drafter_selected)
|
|
|
263 |
|
264 |
class_sizes(db_bb, classes)
|
265 |
circuit_annotations(db_bb, classes)
|
|
|
266 |
class_distribution(db_bb, classes)
|
267 |
-
|
268 |
consistency_circuit(db_bb, classes)
|
269 |
text_statistics(db_bb)
|
|
|
7 |
|
8 |
# Project Imports
|
9 |
from loader import load_classes, load_properties, read_dataset, write_dataset, file_name
|
10 |
+
from utils import bbdist
|
11 |
|
12 |
# Third-Party Imports
|
13 |
import matplotlib.pyplot as plt
|
|
|
30 |
}
|
31 |
|
32 |
|
33 |
+
def consistency(db: list, classes: dict, recover: dict = {}, skip_texts=False) -> tuple:
|
34 |
"""Checks Whether Annotation Classes are in provided Classes Dict and Attempts Recovery"""
|
35 |
|
36 |
total, ok, mapped, faulty, rotation, text = 0, 0, 0, 0, 0, 0
|
37 |
|
38 |
for sample in db:
|
39 |
+
for annotation in sample["bboxes"] + sample["polygons"] + sample["points"]:
|
40 |
total += 1
|
41 |
|
42 |
+
if annotation["class"] in classes:
|
43 |
ok += 1
|
44 |
|
45 |
+
if annotation["class"] in recover:
|
46 |
+
annotation["class"] = recover[annotation["class"]]
|
47 |
mapped += 1
|
48 |
|
49 |
+
if annotation["class"] not in classes and annotation["class"] not in recover:
|
50 |
+
print(f"Can't recover faulty label in {file_name(sample)}: {annotation['class']}")
|
51 |
faulty += 1
|
52 |
|
53 |
+
if annotation["rotation"] is not None:
|
54 |
rotation += 1
|
55 |
|
56 |
+
if not skip_texts:
|
57 |
+
if annotation["class"] == "text" and annotation["text"] is None:
|
58 |
+
print(f"Missing Text in {file_name(sample)} -> {annotation['xmin']}, {annotation['ymin']}")
|
59 |
|
60 |
+
if annotation["text"] is not None:
|
61 |
+
if annotation["text"].strip() != annotation["text"]:
|
62 |
+
print(f"Removing leading of trailing spaces from: {annotation['text']}")
|
63 |
+
annotation["text"] = annotation["text"].strip()
|
64 |
|
65 |
+
if annotation["class"] != "text":
|
66 |
+
print(f"Text string outside Text Annotation in {file_name(sample)} [{annotation['xmin']:4}, {annotation['ymin']:4}]: {annotation['class']}: {annotation['text']}")
|
67 |
|
68 |
+
text += 1
|
69 |
|
70 |
return total, ok, mapped, faulty, rotation, text
|
71 |
|
|
|
111 |
plt.show()
|
112 |
|
113 |
|
114 |
+
def annotation_distribution(db: list) -> None:
|
115 |
+
|
116 |
+
amount_distribution([sample['bboxes'] for sample in db],
|
117 |
+
"Image Sample Count by BB Annotation Count",
|
118 |
+
"BB Annotation Count",
|
119 |
+
"Image Sample Count",
|
120 |
+
ticks=False)
|
121 |
+
|
122 |
+
|
123 |
def class_distribution(db: list, classes: dict) -> None:
|
124 |
"""Plots the Class Distribution over the Dataset"""
|
125 |
|
126 |
class_nbrs = np.arange(len(classes))
|
127 |
+
class_counts = [sum([len([annotation for annotation in sample["bboxes"] + sample["polygons"] + sample["points"]
|
128 |
+
if annotation["class"] == cls])
|
129 |
for sample in db]) for cls in classes]
|
130 |
|
131 |
bars = plt.bar(class_nbrs, class_counts)
|
|
|
165 |
(not drafter or f"drafter_{drafter}{os.sep}" in root)])
|
166 |
|
167 |
|
168 |
+
def read_check_write(classes: dict, drafter: int = None, segmentation: bool = False) -> list:
|
169 |
"""Reads Annotations, Checks Consistency with Provided Classes
|
170 |
Writes Corrected Annotations Back and Returns the Annotations"""
|
171 |
|
172 |
db = read_dataset(drafter=drafter, segmentation=segmentation)
|
173 |
+
ann_total, ann_ok, ann_mapped, ann_faulty, ann_rot, ann_text = consistency(db,
|
174 |
+
classes,
|
175 |
+
MAPPING_LOOKUP,
|
176 |
+
skip_texts=segmentation)
|
177 |
write_dataset(db, segmentation=segmentation)
|
178 |
|
179 |
print("")
|
|
|
193 |
return db
|
194 |
|
195 |
|
196 |
+
def unique_characters(texts: list) -> list:
|
197 |
+
"""Returns the Sorted Set of Unique Characters"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
198 |
|
199 |
+
char_set = set([char for text in texts for char in text])
|
200 |
+
return sorted(list(char_set))
|
201 |
|
|
|
202 |
|
203 |
+
def character_distribution(texts: list, chars: list):
|
204 |
+
"""Plots and Returns the Character Distribution"""
|
|
|
|
|
205 |
|
206 |
char_nbrs = np.arange(len(chars))
|
207 |
char_counts = [sum([len([None for text_char in text_label if text_char == char])
|
208 |
+
for text_label in texts])
|
209 |
for char in chars]
|
210 |
plt.bar(char_nbrs, char_counts)
|
211 |
plt.xticks(char_nbrs, chars)
|
212 |
+
plt.title("Character Distribution")
|
213 |
plt.xlabel("Character")
|
214 |
plt.ylabel("Overall Count")
|
215 |
plt.show()
|
216 |
+
|
217 |
+
return char_counts
|
218 |
+
|
219 |
+
|
220 |
+
def amount_distribution(list_of_lists: list, title: str, x_label: str, y_label: str, ticks: bool = True) -> None:
|
221 |
+
"""Plots a Histogram of the Amount of Things Contained in a List of Lists"""
|
222 |
+
|
223 |
+
max_bin = max([len(lst) for lst in list_of_lists])
|
224 |
+
bin_numbers = np.arange(max_bin)+1
|
225 |
+
text_count_by_length = [len([None for lst in list_of_lists if len(lst) == amount])
|
226 |
+
for amount in bin_numbers]
|
227 |
+
plt.bar(bin_numbers, text_count_by_length)
|
228 |
+
|
229 |
+
if ticks:
|
230 |
+
plt.xticks(bin_numbers, rotation=90)
|
231 |
+
|
232 |
+
plt.title(title)
|
233 |
+
plt.xlabel(x_label)
|
234 |
+
plt.ylabel(y_label)
|
235 |
+
plt.show()
|
236 |
+
|
237 |
+
|
238 |
+
def text_proximity(db: list, cls_name: str, cls_regex: str):
|
239 |
+
"""Proximity-Based Regex Validation"""
|
240 |
+
|
241 |
+
cls_stat = {}
|
242 |
+
|
243 |
+
for sample in db:
|
244 |
+
bbs_text = [bbox for bbox in sample["bboxes"] if bbox["class"] == "text"]
|
245 |
+
bbs_symbol = [bbox for bbox in sample["bboxes"] if bbox["class"] not in ["text", "junction", "crossover"]]
|
246 |
+
|
247 |
+
for bb_text in bbs_text:
|
248 |
+
if bb_text["text"]:
|
249 |
+
if re.match(cls_regex, bb_text["text"]):
|
250 |
+
bb_closest_class = sorted(bbs_symbol, key=lambda bb: bbdist(bb_text, bb))[0]["class"]
|
251 |
+
cls_stat[bb_closest_class] = cls_stat.get(bb_closest_class, 0) + 1
|
252 |
+
|
253 |
+
cls_stat = sorted(cls_stat.items(), key=lambda cls: -cls[1])
|
254 |
+
print(cls_stat)
|
255 |
+
plt.bar(range(len(cls_stat)), [name for _, name in cls_stat])
|
256 |
+
plt.xticks(range(len(cls_stat)), labels=[name for name, _ in cls_stat], rotation=90)
|
257 |
+
plt.title(f"Neighbor Distribution for {cls_name} Text Annotations")
|
258 |
+
plt.xlabel("Symbol Class")
|
259 |
+
plt.ylabel("Number of Closest Neighbors")
|
260 |
+
plt.tight_layout()
|
261 |
plt.show()
|
262 |
|
263 |
+
|
264 |
+
def text_statistics(db: list, plot_unique_labels: bool = False):
|
265 |
+
"""Generates and Plots Statistics on Text Classes"""
|
266 |
+
|
267 |
+
text_bbs = [bbox for sample in db for bbox in sample["bboxes"] if bbox["class"] == "text"]
|
268 |
+
text_labels = [bbox["text"] for bbox in text_bbs if type(bbox["text"]) is str and len(text_bbs) > 0]
|
269 |
+
text_labels_unique = set(text_labels)
|
270 |
+
chars_unique = unique_characters(text_labels)
|
271 |
+
char_counts = character_distribution(text_labels, chars_unique)
|
272 |
+
amount_distribution(text_labels, "Text Length Distribution", "Character Count", "Annotation Count")
|
273 |
+
|
274 |
+
print("")
|
275 |
+
print(" Text Statistics")
|
276 |
+
print("---------------------")
|
277 |
+
print(f"Text BB Annotations: {len(text_bbs)}")
|
278 |
+
print(f"Overall Text Label Count: {len(text_labels)}")
|
279 |
+
print(f"Annotation Completeness: {100*len(text_labels)/len(text_bbs):.2f}%")
|
280 |
+
print(f"Unique Text Label Count: {len(text_labels_unique)}")
|
281 |
+
print(f"Total Character Count: {sum([len(text_label) for text_label in text_labels])}")
|
282 |
+
print(f"Character Types: {len(chars_unique)}")
|
283 |
+
print("\n\nSet of all characters occurring in all text labels:")
|
284 |
+
print(chars_unique)
|
285 |
+
print("\n\nSet of Text Labels:")
|
286 |
+
print(text_labels_unique)
|
287 |
+
|
288 |
+
print("\nCharacter Frequencies:")
|
289 |
+
print({char: 1/char_count for char, char_count in zip(chars_unique, char_counts)})
|
290 |
+
|
291 |
text_instances = text_labels_unique if plot_unique_labels else text_labels
|
292 |
text_classes_names = []
|
293 |
text_classes_instances = []
|
|
|
314 |
plt.tight_layout()
|
315 |
plt.show()
|
316 |
|
317 |
+
text_proximity(db, "Capacitor Name", "^C[0-9]+$")
|
318 |
+
text_proximity(db, "Resistor Name", "^R[0-9]+$")
|
319 |
+
text_proximity(db, "Inductor Name", "^L[0-9]+$")
|
320 |
+
|
321 |
|
322 |
|
323 |
if __name__ == "__main__":
|
324 |
drafter_selected = int(sys.argv[1]) if len(sys.argv) == 2 else None
|
325 |
classes = load_classes()
|
326 |
|
|
|
327 |
db_bb = read_check_write(classes, drafter_selected)
|
328 |
+
db_poly = read_check_write(classes, drafter_selected, segmentation=True)
|
329 |
|
330 |
class_sizes(db_bb, classes)
|
331 |
circuit_annotations(db_bb, classes)
|
332 |
+
annotation_distribution(db_bb)
|
333 |
class_distribution(db_bb, classes)
|
334 |
+
class_distribution(db_poly, classes)
|
335 |
consistency_circuit(db_bb, classes)
|
336 |
text_statistics(db_bb)
|
loader.py
CHANGED
@@ -49,10 +49,16 @@ def _sample_info_from_path(path: str) -> tuple:
|
|
49 |
return drafter.split("_")[1], int(circuit[1:]), int(drawing[1:]), int(picture[1:]), suffix
|
50 |
|
51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
def file_name(sample: dict) -> str:
|
53 |
"""return the Raw Image File Name of a Sample"""
|
54 |
|
55 |
-
return f"
|
56 |
|
57 |
|
58 |
def read_pascal_voc(path: str) -> dict:
|
@@ -127,10 +133,15 @@ def read_labelme(path: str) -> dict:
|
|
127 |
drafter, circuit, drawing, picture, _ = _sample_info_from_path(path)
|
128 |
suffix = json_data['imagePath'].split(".")[-1]
|
129 |
|
130 |
-
return {'img_path': json_data['imagePath'], 'drafter': drafter, 'circuit': circuit,
|
131 |
'drawing': drawing, 'picture': picture, 'format': suffix,
|
132 |
'height': json_data['imageHeight'], 'width': json_data['imageWidth'], 'bboxes': [],
|
133 |
-
'polygons': [{'class': shape['label'],
|
|
|
|
|
|
|
|
|
|
|
134 |
'rotation': shape.get('rotation', None),
|
135 |
'text': shape.get('text', None),
|
136 |
'group': shape.get('group_id', None)}
|
@@ -150,19 +161,25 @@ def write_labelme(geo_data: dict, path: str = None) -> None:
|
|
150 |
f"C{geo_data['circuit']}_D{geo_data['drawing']}_P{geo_data['picture']}.json")
|
151 |
|
152 |
with open(path, 'w') as json_file:
|
153 |
-
json.dump({'version': '
|
154 |
-
'
|
155 |
-
'
|
156 |
-
'shapes': [{'label': polygon['class'], 'line_color': None, 'fill_color': None,
|
157 |
'points': polygon['points'],
|
158 |
-
|
|
|
159 |
**({'rotation': polygon['rotation']} if polygon.get('rotation', None) else {}),
|
160 |
**({'text': polygon['text']} if polygon.get('text', None) else {}),
|
161 |
'shape_type': 'polygon', 'flags': {}}
|
162 |
for polygon in geo_data['polygons']] +
|
163 |
[{'label': point['class'], 'points': [[point['points'][0], point['points'][1]]],
|
164 |
'group_id': point['group'], 'shape_type': 'point', 'flags': {}}
|
165 |
-
for point in geo_data['points']]
|
|
|
|
|
|
|
|
|
|
|
|
|
166 |
json_file, indent=2)
|
167 |
|
168 |
|
@@ -206,7 +223,7 @@ def read_snippets(**kwargs):
|
|
206 |
|
207 |
for img, annotations in read_images(**kwargs):
|
208 |
for bbox in annotations['bboxes']:
|
209 |
-
snippets += [(img[bbox['ymin']:bbox['ymax'], bbox['xmin']:bbox['xmax']], bbox)]
|
210 |
|
211 |
return snippets
|
212 |
|
@@ -217,11 +234,11 @@ if __name__ == "__main__":
|
|
217 |
os.mkdir("test")
|
218 |
args = {'drafter': int(sys.argv[1])} if len(sys.argv) == 2 else {}
|
219 |
|
220 |
-
for
|
221 |
if bbox['class'] == "text" and bbox.get("text", ""):
|
222 |
if bbox['rotation'] == 90:
|
223 |
snippet = cv2.rotate(snippet, cv2.ROTATE_90_CLOCKWISE)
|
224 |
if bbox['rotation'] == 270:
|
225 |
snippet = cv2.rotate(snippet, cv2.ROTATE_90_COUNTERCLOCKWISE)
|
226 |
|
227 |
-
cv2.imwrite(join("test", f"{bbox['text']}
|
|
|
49 |
return drafter.split("_")[1], int(circuit[1:]), int(drawing[1:]), int(picture[1:]), suffix
|
50 |
|
51 |
|
52 |
+
def sample_name(sample: dict) -> str:
|
53 |
+
"""Returns the Name of a Sample"""
|
54 |
+
|
55 |
+
return f"C{sample['circuit']}_D{sample['drawing']}_P{sample['picture']}"
|
56 |
+
|
57 |
+
|
58 |
def file_name(sample: dict) -> str:
|
59 |
"""return the Raw Image File Name of a Sample"""
|
60 |
|
61 |
+
return f"{sample_name(sample)}.{sample['format']}"
|
62 |
|
63 |
|
64 |
def read_pascal_voc(path: str) -> dict:
|
|
|
133 |
drafter, circuit, drawing, picture, _ = _sample_info_from_path(path)
|
134 |
suffix = json_data['imagePath'].split(".")[-1]
|
135 |
|
136 |
+
return {'img_path': json_data['imagePath'].replace("\\", "/"), 'drafter': drafter, 'circuit': circuit,
|
137 |
'drawing': drawing, 'picture': picture, 'format': suffix,
|
138 |
'height': json_data['imageHeight'], 'width': json_data['imageWidth'], 'bboxes': [],
|
139 |
+
'polygons': [{'class': shape['label'],
|
140 |
+
'bbox': {'xmin': min(point[0] for point in shape['points']),
|
141 |
+
'ymin': min(point[1] for point in shape['points']),
|
142 |
+
'xmax': max(point[0] for point in shape['points']),
|
143 |
+
'ymax': max(point[1] for point in shape['points'])},
|
144 |
+
'points': shape['points'],
|
145 |
'rotation': shape.get('rotation', None),
|
146 |
'text': shape.get('text', None),
|
147 |
'group': shape.get('group_id', None)}
|
|
|
161 |
f"C{geo_data['circuit']}_D{geo_data['drawing']}_P{geo_data['picture']}.json")
|
162 |
|
163 |
with open(path, 'w') as json_file:
|
164 |
+
json.dump({'version': '5.2.0',
|
165 |
+
'flags': {},
|
166 |
+
'shapes': [{'line_color': None, 'fill_color': None,'label': polygon['class'],
|
|
|
167 |
'points': polygon['points'],
|
168 |
+
'group_id': polygon.get('group', None),
|
169 |
+
'description': polygon.get('description', None),
|
170 |
**({'rotation': polygon['rotation']} if polygon.get('rotation', None) else {}),
|
171 |
**({'text': polygon['text']} if polygon.get('text', None) else {}),
|
172 |
'shape_type': 'polygon', 'flags': {}}
|
173 |
for polygon in geo_data['polygons']] +
|
174 |
[{'label': point['class'], 'points': [[point['points'][0], point['points'][1]]],
|
175 |
'group_id': point['group'], 'shape_type': 'point', 'flags': {}}
|
176 |
+
for point in geo_data['points']],
|
177 |
+
'imagePath': geo_data['img_path'],
|
178 |
+
'imageData': None,
|
179 |
+
'imageHeight': geo_data['height'],
|
180 |
+
'imageWidth': geo_data['width'],
|
181 |
+
'lineColor': [0, 255, 0, 128],
|
182 |
+
'fillColor': [255, 0, 0, 128]},
|
183 |
json_file, indent=2)
|
184 |
|
185 |
|
|
|
223 |
|
224 |
for img, annotations in read_images(**kwargs):
|
225 |
for bbox in annotations['bboxes']:
|
226 |
+
snippets += [(img[bbox['ymin']:bbox['ymax'], bbox['xmin']:bbox['xmax']], bbox, sample_name(annotations))]
|
227 |
|
228 |
return snippets
|
229 |
|
|
|
234 |
os.mkdir("test")
|
235 |
args = {'drafter': int(sys.argv[1])} if len(sys.argv) == 2 else {}
|
236 |
|
237 |
+
for snippet, bbox, sample in read_snippets(**args):
|
238 |
if bbox['class'] == "text" and bbox.get("text", ""):
|
239 |
if bbox['rotation'] == 90:
|
240 |
snippet = cv2.rotate(snippet, cv2.ROTATE_90_CLOCKWISE)
|
241 |
if bbox['rotation'] == 270:
|
242 |
snippet = cv2.rotate(snippet, cv2.ROTATE_90_COUNTERCLOCKWISE)
|
243 |
|
244 |
+
cv2.imwrite(join("test", f"{bbox['text']}___{sample}_{bbox['ymin']}_{bbox['ymax']}_{bbox['xmin']}_{bbox['xmax']}.png"), snippet)
|
requirements.txt
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
contourpy==1.1.0
|
2 |
+
cv==1.0.0
|
3 |
+
cycler==0.11.0
|
4 |
+
fonttools==4.40.0
|
5 |
+
kiwisolver==1.4.4
|
6 |
+
lxml==4.9.2
|
7 |
+
matplotlib==3.7.1
|
8 |
+
numpy==1.25.0
|
9 |
+
opencv-python==4.7.0.72
|
10 |
+
packaging==23.1
|
11 |
+
Pillow==9.5.0
|
12 |
+
pyparsing==3.1.0
|
13 |
+
python-dateutil==2.8.2
|
14 |
+
six==1.16.0
|
segmentation.py
CHANGED
@@ -9,7 +9,7 @@ from math import dist
|
|
9 |
|
10 |
# Project Imports
|
11 |
from loader import read_pascal_voc, read_labelme, write_labelme, load_classes_ports
|
12 |
-
from utils import transform, associated_keypoints
|
13 |
|
14 |
# Third-Party Imports
|
15 |
import cv2
|
@@ -34,9 +34,8 @@ def binary_to_multi_seg_map(drafter: str, sample: str, suffix: str, source_folde
|
|
34 |
shape_mask = np.ones(bin_seg_map.shape, dtype=np.uint8)*255
|
35 |
geo_data = read_labelme(join(drafter, source_folder, f"{sample}.json"))
|
36 |
|
37 |
-
for shape in sorted(geo_data["polygons"],
|
38 |
-
|
39 |
-
(max([p[1] for p in shape['points']])-min([p[1] for p in shape['points']]))):
|
40 |
cv2.fillPoly(shape_mask,
|
41 |
pts=[np.array(shape["points"], dtype=np.int32)],
|
42 |
color=color_map[shape["class"]])
|
@@ -80,10 +79,10 @@ def generate_keypoints(drafter: str, sample: str, suffix: str, source_folder: st
|
|
80 |
shape['group'] = nbr
|
81 |
|
82 |
if shape['class'] != "text" and shape['class'] != "wire":
|
83 |
-
x_min = max(int(
|
84 |
-
x_max = min(int(
|
85 |
-
y_min = max(int(
|
86 |
-
y_max = min(int(
|
87 |
cropout = bin_seg_map[y_min:y_max, x_min:x_max]
|
88 |
shape_mask = np.zeros((y_max-y_min, x_max-x_min), dtype=np.uint8)
|
89 |
cv2.polylines(shape_mask, pts=[np.array(shape["points"]-np.array([[x_min, y_min]]), dtype=np.int32)],
|
@@ -210,25 +209,6 @@ def refine_polygons(drafter: str, sample: str, suffix: str, source_folder: str,
|
|
210 |
write_labelme(geo_data, join(drafter, target_folder, f"{sample}.json"))
|
211 |
|
212 |
|
213 |
-
def bounding_box(points):
|
214 |
-
xmin = min(point[0] for point in points)
|
215 |
-
ymin = min(point[1] for point in points)
|
216 |
-
xmax = max(point[0] for point in points)
|
217 |
-
ymax = max(point[1] for point in points)
|
218 |
-
return [xmin, ymin, xmax, ymax]
|
219 |
-
|
220 |
-
|
221 |
-
def overlap(bbox1, bbox2):
|
222 |
-
|
223 |
-
if bbox1["xmin"] > bbox2[2] or bbox1["xmax"] < bbox2[0]:
|
224 |
-
return False
|
225 |
-
|
226 |
-
if bbox1["ymin"] > bbox2[3] or bbox1["ymax"] < bbox2[1]:
|
227 |
-
return False
|
228 |
-
|
229 |
-
return True
|
230 |
-
|
231 |
-
|
232 |
def find_closest_points(list1, list2):
|
233 |
reordered_list2 = []
|
234 |
for x1, y1 in list1:
|
@@ -256,7 +236,7 @@ def connector_type_assignment(drafter: str, sample: str, suffix: str, source_fol
|
|
256 |
connectors = associated_keypoints(instances, shape)
|
257 |
cls_ports = classes_ports[shape["class"]]
|
258 |
bboxes_match = [bbox for bbox in bboxes['bboxes']
|
259 |
-
if overlap(bbox,
|
260 |
|
261 |
if len(cls_ports) != len(connectors):
|
262 |
print(f" Bad Connector Count: {shape['class']} {shape['points'][0]} -> {len(cls_ports)} vs. {len(connectors)}")
|
|
|
9 |
|
10 |
# Project Imports
|
11 |
from loader import read_pascal_voc, read_labelme, write_labelme, load_classes_ports
|
12 |
+
from utils import transform, associated_keypoints, overlap
|
13 |
|
14 |
# Third-Party Imports
|
15 |
import cv2
|
|
|
34 |
shape_mask = np.ones(bin_seg_map.shape, dtype=np.uint8)*255
|
35 |
geo_data = read_labelme(join(drafter, source_folder, f"{sample}.json"))
|
36 |
|
37 |
+
for shape in sorted(geo_data["polygons"], key=lambda shape: -(shape['bbox']['xmax']-shape['bbox']['xmin']) *
|
38 |
+
(shape['bbox']['ymax']-shape['bbox']['ymin'])):
|
|
|
39 |
cv2.fillPoly(shape_mask,
|
40 |
pts=[np.array(shape["points"], dtype=np.int32)],
|
41 |
color=color_map[shape["class"]])
|
|
|
79 |
shape['group'] = nbr
|
80 |
|
81 |
if shape['class'] != "text" and shape['class'] != "wire":
|
82 |
+
x_min = max(int(shape['bbox']['xmin'])-margin, 0)
|
83 |
+
x_max = min(int(shape['bbox']['xmax'])+margin, bin_seg_map.shape[1])
|
84 |
+
y_min = max(int(shape['bbox']['ymin'])-margin, 0)
|
85 |
+
y_max = min(int(shape['bbox']['ymax'])+margin, bin_seg_map.shape[0])
|
86 |
cropout = bin_seg_map[y_min:y_max, x_min:x_max]
|
87 |
shape_mask = np.zeros((y_max-y_min, x_max-x_min), dtype=np.uint8)
|
88 |
cv2.polylines(shape_mask, pts=[np.array(shape["points"]-np.array([[x_min, y_min]]), dtype=np.int32)],
|
|
|
209 |
write_labelme(geo_data, join(drafter, target_folder, f"{sample}.json"))
|
210 |
|
211 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
212 |
def find_closest_points(list1, list2):
|
213 |
reordered_list2 = []
|
214 |
for x1, y1 in list1:
|
|
|
236 |
connectors = associated_keypoints(instances, shape)
|
237 |
cls_ports = classes_ports[shape["class"]]
|
238 |
bboxes_match = [bbox for bbox in bboxes['bboxes']
|
239 |
+
if overlap(bbox, shape["bbox"]) and bbox['class'] == shape['class']]
|
240 |
|
241 |
if len(cls_ports) != len(connectors):
|
242 |
print(f" Bad Connector Count: {shape['class']} {shape['points'][0]} -> {len(cls_ports)} vs. {len(connectors)}")
|
utils.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
"""utils.py: Helper Functions to keep this Repo Standalone"""
|
2 |
|
3 |
# System Imports
|
4 |
-
from math import sin, cos, radians
|
5 |
|
6 |
__author__ = "Johannes Bayer"
|
7 |
__copyright__ = "Copyright 2023, DFKI"
|
@@ -42,6 +42,24 @@ def transform(port, bb):
|
|
42 |
return {"name": port['name'], "position": p}
|
43 |
|
44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
def associated_keypoints(instances, shape):
|
46 |
"""Returns the points with same group id as the provided polygon"""
|
47 |
|
@@ -49,9 +67,6 @@ def associated_keypoints(instances, shape):
|
|
49 |
if point["group"] == shape["group"] and point["class"] == "connector"]
|
50 |
|
51 |
|
52 |
-
def poly_to_bb():
|
53 |
-
pass
|
54 |
-
|
55 |
def IoU(bb1, bb2):
|
56 |
"""Intersection over Union"""
|
57 |
|
|
|
1 |
"""utils.py: Helper Functions to keep this Repo Standalone"""
|
2 |
|
3 |
# System Imports
|
4 |
+
from math import sin, cos, radians, sqrt
|
5 |
|
6 |
__author__ = "Johannes Bayer"
|
7 |
__copyright__ = "Copyright 2023, DFKI"
|
|
|
42 |
return {"name": port['name'], "position": p}
|
43 |
|
44 |
|
45 |
+
def bbdist(bb1, bb2):
|
46 |
+
"""Calculates the Distance between two Bounding Box Annotations"""
|
47 |
+
|
48 |
+
return sqrt(((bb1["xmin"]+bb1["xmax"])/2 - (bb2["xmin"]+bb2["xmax"])/2)**2 +
|
49 |
+
((bb1["ymin"]+bb1["ymax"])/2 - (bb2["ymin"]+bb2["ymax"])/2)**2)
|
50 |
+
|
51 |
+
|
52 |
+
def overlap(bbox1, bbox2):
|
53 |
+
|
54 |
+
if bbox1["xmin"] > bbox2["xmax"] or bbox1["xmax"] < bbox2["xmin"]:
|
55 |
+
return False
|
56 |
+
|
57 |
+
if bbox1["ymin"] > bbox2["ymax"] or bbox1["ymax"] < bbox2["ymin"]:
|
58 |
+
return False
|
59 |
+
|
60 |
+
return True
|
61 |
+
|
62 |
+
|
63 |
def associated_keypoints(instances, shape):
|
64 |
"""Returns the points with same group id as the provided polygon"""
|
65 |
|
|
|
67 |
if point["group"] == shape["group"] and point["class"] == "connector"]
|
68 |
|
69 |
|
|
|
|
|
|
|
70 |
def IoU(bb1, bb2):
|
71 |
"""Intersection over Union"""
|
72 |
|