-
Notifications
You must be signed in to change notification settings - Fork 7
/
eval.py
385 lines (308 loc) · 11.8 KB
/
eval.py
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
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
# -*- coding: utf-8 -*-
import argparse
import os
import cv2
import numpy as np
import torch
import torch.backends.cudnn as cudnn
from tqdm import tqdm
import wandb
from config.load_config import load_yaml, DotDict
from model.craft import CRAFT
from metrics.eval_det_iou import DetectionIoUEvaluator
from utils.inference_boxes import (
test_net,
load_icdar2015_gt,
load_icdar2013_gt,
load_synthtext_gt,
)
from utils.util import copyStateDict
def save_result_synth(img_file, img, pre_output, pre_box, gt_box=None, result_dir=""):
img = np.array(img)
img_copy = img.copy()
region = pre_output[0]
affinity = pre_output[1]
# make result file list
filename, file_ext = os.path.splitext(os.path.basename(img_file))
# draw bounding boxes for prediction, color green
for i, box in enumerate(pre_box):
poly = np.array(box).astype(np.int32).reshape((-1))
poly = poly.reshape(-1, 2)
try:
cv2.polylines(
img, [poly.reshape((-1, 1, 2))], True, color=(0, 255, 0), thickness=2
)
except:
pass
# draw bounding boxes for gt, color red
if gt_box is not None:
for j in range(len(gt_box)):
cv2.polylines(
img,
[np.array(gt_box[j]["points"]).astype(np.int32).reshape((-1, 1, 2))],
True,
color=(0, 0, 255),
thickness=2,
)
# draw overlay image
overlay_img = overlay(img_copy, region, affinity, pre_box)
# Save result image
res_img_path = result_dir + "/res_" + filename + ".jpg"
cv2.imwrite(res_img_path, img)
overlay_image_path = result_dir + "/res_" + filename + "_box.jpg"
cv2.imwrite(overlay_image_path, overlay_img)
def save_result_2015(img_file, img, pre_output, pre_box, gt_box, result_dir):
img = np.array(img)
img_copy = img.copy()
region = pre_output[0]
affinity = pre_output[1]
# make result file list
filename, file_ext = os.path.splitext(os.path.basename(img_file))
for i, box in enumerate(pre_box):
poly = np.array(box).astype(np.int32).reshape((-1))
poly = poly.reshape(-1, 2)
try:
cv2.polylines(
img, [poly.reshape((-1, 1, 2))], True, color=(0, 255, 0), thickness=2
)
except:
pass
if gt_box is not None:
for j in range(len(gt_box)):
_gt_box = np.array(gt_box[j]["points"]).reshape(-1, 2).astype(np.int32)
if gt_box[j]["text"] == "###":
cv2.polylines(img, [_gt_box], True, color=(128, 128, 128), thickness=2)
else:
cv2.polylines(img, [_gt_box], True, color=(0, 0, 255), thickness=2)
# draw overlay image
overlay_img = overlay(img_copy, region, affinity, pre_box)
# Save result image
res_img_path = result_dir + "/res_" + filename + ".jpg"
cv2.imwrite(res_img_path, img)
overlay_image_path = result_dir + "/res_" + filename + "_box.jpg"
cv2.imwrite(overlay_image_path, overlay_img)
def save_result_2013(img_file, img, pre_output, pre_box, gt_box=None, result_dir=""):
img = np.array(img)
img_copy = img.copy()
region = pre_output[0]
affinity = pre_output[1]
# make result file list
filename, file_ext = os.path.splitext(os.path.basename(img_file))
# draw bounding boxes for prediction, color green
for i, box in enumerate(pre_box):
poly = np.array(box).astype(np.int32).reshape((-1))
poly = poly.reshape(-1, 2)
try:
cv2.polylines(
img, [poly.reshape((-1, 1, 2))], True, color=(0, 255, 0), thickness=2
)
except:
pass
# draw bounding boxes for gt, color red
if gt_box is not None:
for j in range(len(gt_box)):
cv2.polylines(
img,
[np.array(gt_box[j]["points"]).reshape((-1, 1, 2))],
True,
color=(0, 0, 255),
thickness=2,
)
# draw overlay image
overlay_img = overlay(img_copy, region, affinity, pre_box)
# Save result image
res_img_path = result_dir + "/res_" + filename + ".jpg"
cv2.imwrite(res_img_path, img)
overlay_image_path = result_dir + "/res_" + filename + "_box.jpg"
cv2.imwrite(overlay_image_path, overlay_img)
def overlay(image, region, affinity, single_img_bbox):
height, width, channel = image.shape
region_score = cv2.resize(region, (width, height))
affinity_score = cv2.resize(affinity, (width, height))
overlay_region = cv2.addWeighted(image.copy(), 0.4, region_score, 0.6, 5)
overlay_aff = cv2.addWeighted(image.copy(), 0.4, affinity_score, 0.6, 5)
boxed_img = image.copy()
for word_box in single_img_bbox:
cv2.polylines(
boxed_img,
[word_box.astype(np.int32).reshape((-1, 1, 2))],
True,
color=(0, 255, 0),
thickness=3,
)
temp1 = np.hstack([image, boxed_img])
temp2 = np.hstack([overlay_region, overlay_aff])
temp3 = np.vstack([temp1, temp2])
return temp3
def load_test_dataset_iou(test_folder_name, config):
if test_folder_name == "synthtext":
total_bboxes_gt, total_img_path = load_synthtext_gt(config.test_data_dir)
elif test_folder_name == "icdar2013":
total_bboxes_gt, total_img_path = load_icdar2013_gt(
dataFolder=config.test_data_dir
)
elif test_folder_name == "icdar2015":
total_bboxes_gt, total_img_path = load_icdar2015_gt(
dataFolder=config.test_data_dir
)
else:
print("not found test dataset")
return None, None
return total_bboxes_gt, total_img_path
def viz_test(img, pre_output, pre_box, gt_box, img_name, result_dir, test_folder_name):
if test_folder_name == "synthtext":
save_result_synth(
img_name, img[:, :, ::-1].copy(), pre_output, pre_box, gt_box, result_dir
)
elif test_folder_name == "icdar2013":
save_result_2013(
img_name, img[:, :, ::-1].copy(), pre_output, pre_box, gt_box, result_dir
)
elif test_folder_name == "icdar2015":
save_result_2015(
img_name, img[:, :, ::-1].copy(), pre_output, pre_box, gt_box, result_dir
)
else:
print("not found test dataset")
def load_gt_cl_dir(config, data):
if data == "icdar2013":
gt_cl_dir = os.path.join(config.test_data_dir, "Challenge2_Test_Task1_GT_cl")
elif data == "icdar2015":
gt_cl_dir = os.path.join(
config.test_data_dir, "ch4_test_localization_transcription_gt_cl"
)
else:
gt_cl_dir = None
print("no dataset")
return gt_cl_dir
def main_eval(model_path, backbone, config, evaluator, result_dir, buffer, model, mode):
if not os.path.exists(result_dir):
os.makedirs(result_dir, exist_ok=True)
test_set = config.test_data_dir.split("/")[-2].lower()
total_imgs_bboxes_gt, total_imgs_path = load_test_dataset_iou(test_set, config)
if mode == "weak_supervision":
gpu_count = torch.cuda.device_count() // 2
else:
gpu_count = torch.cuda.device_count()
gpu_idx = torch.cuda.current_device()
torch.cuda.set_device(gpu_idx)
# Only evaluation time
if model is None:
piece_imgs_path = total_imgs_path
if backbone == "vgg":
model = CRAFT()
else:
raise Exception("Undefined architecture")
print("Loading weights from checkpoint (" + model_path + ")")
net_param = torch.load(model_path, map_location=f"cuda:{gpu_idx}")
model.load_state_dict(copyStateDict(net_param["craft"]))
if config.cuda:
model = model.cuda()
cudnn.benchmark = False
# Distributed evaluation in the middle of training time
else:
# check all buffer value is None for distributed evaluation
assert all(
v is None for v in buffer
), "Buffer already filled with another value."
slice_idx = len(total_imgs_bboxes_gt) // gpu_count
# last gpu
if gpu_idx == gpu_count - 1:
piece_imgs_path = total_imgs_path[gpu_idx * slice_idx :]
# piece_imgs_bboxes_gt = total_imgs_bboxes_gt[gpu_idx * slice_idx:]
else:
piece_imgs_path = total_imgs_path[
gpu_idx * slice_idx : (gpu_idx + 1) * slice_idx
]
# piece_imgs_bboxes_gt = total_imgs_bboxes_gt[gpu_idx * slice_idx: (gpu_idx + 1) * slice_idx]
model.eval()
# -----------------------------------------------------------------------------------------------------------------#
total_imgs_bboxes_pre = []
for k, img_path in enumerate(tqdm(piece_imgs_path)):
image = cv2.imread(img_path)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
single_img_bbox = []
bboxes, polys, score_text = test_net(
model,
image,
config.text_threshold,
config.link_threshold,
config.low_text,
config.cuda,
config.poly,
config.canvas_size,
config.mag_ratio,
)
for box in bboxes:
box_info = {"points": box, "text": "###", "ignore": False}
single_img_bbox.append(box_info)
total_imgs_bboxes_pre.append(single_img_bbox)
# Distributed evaluation -------------------------------------------------------------------------------------#
if buffer is not None:
buffer[gpu_idx * slice_idx + k] = single_img_bbox
# print(sum([element is not None for element in buffer]))
# -------------------------------------------------------------------------------------------------------------#
if config.vis_opt:
viz_test(
image,
score_text,
pre_box=polys,
gt_box=total_imgs_bboxes_gt[k],
img_name=img_path,
result_dir=result_dir,
test_folder_name=test_set,
)
# When distributed evaluation mode, wait until buffer is full filled
if buffer is not None:
while None in buffer:
continue
assert all(v is not None for v in buffer), "Buffer not filled"
total_imgs_bboxes_pre = buffer
results = []
for i, (gt, pred) in enumerate(zip(total_imgs_bboxes_gt, total_imgs_bboxes_pre)):
perSampleMetrics_dict = evaluator.evaluate_image(gt, pred)
results.append(perSampleMetrics_dict)
metrics = evaluator.combine_results(results)
print(metrics)
return metrics
def cal_eval(config, data, res_dir_name, opt, mode):
evaluator = DetectionIoUEvaluator()
test_config = DotDict(config.test[data])
res_dir = os.path.join(os.path.join("exp", args.yaml), "{}".format(res_dir_name))
if opt == "iou_eval":
main_eval(
config.test.trained_model,
config.train.backbone,
test_config,
evaluator,
res_dir,
buffer=None,
model=None,
mode=mode,
)
else:
print("Undefined evaluation")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="CRAFT Text Detection Eval")
parser.add_argument(
"--yaml",
"--yaml_file_name",
default="ic15_train_report4",
type=str,
help="Load configuration",
)
args = parser.parse_args()
# load configure
config = load_yaml(args.yaml)
config = DotDict(config)
if config["wandb_opt"]:
wandb.init(project="evaluation", entity="gmuffiness", name=args.yaml)
wandb.config.update(config)
val_result_dir_name = args.yaml
cal_eval(
config,
"icdar2015",
val_result_dir_name + "-ic15-iou",
opt="iou_eval",
mode=None,
)