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
| import codecs
import os
import time
import sys
sys.path.append('PaddleDetection')
import json
import yaml
from functools import reduce
import multiprocessing
from PIL import Image
import cv2
import numpy as np
import paddle
from paddle.inference import Config
from paddle.inference import create_predictor
from multiprocessing.dummy import Pool as ThreadPool
from functools import partial
from deploy.python.preprocess import preprocess,Resize, NormalizeImage, Permute, PadStride
from deploy.python.utils import argsparser, Timer, get_current_memory_mb
LABEL_MAP = {
"0": "bump",
"1": "granary",
"2": "CrossWalk",
"3": "cone",
"4": "bridge",
"5": "pig",
"6": "tractor",
"7": "corn",
}
class PredictConfig():
def __init__(self, model_dir):
deploy_file = os.path.join(model_dir, 'infer_cfg.yml')
with open(deploy_file) as f:
yml_conf = yaml.safe_load(f)
self.arch = yml_conf['arch']
self.preprocess_infos = yml_conf['Preprocess']
self.min_subgraph_size = yml_conf['min_subgraph_size']
self.labels = yml_conf['label_list']
def get_test_images(infer_file):
with open(infer_file, 'r') as f:
dirs = f.readlines()
images = []
for dir in dirs:
images.append(eval(repr(dir.replace('\n',''))).replace('\\', '/'))
assert len(images) > 0, "no image found in {}".format(infer_file)
return images
def load_predictor(model_dir):
config = Config(
os.path.join(model_dir, 'model.pdmodel'),
os.path.join(model_dir, 'model.pdiparams'))
config.enable_use_gpu(3000, 0)
config.switch_ir_optim(True)
config.disable_glog_info()
config.enable_memory_optim()
config.switch_use_feed_fetch_ops(False)
predictor = create_predictor(config)
return predictor, config
def create_inputs(imgs, im_info):
inputs = {}
im_shape = []
scale_factor = []
if len(imgs) == 1:
inputs['image'] = np.array((imgs[0], )).astype('float32')
inputs['im_shape'] = np.array(
(im_info[0]['im_shape'], )).astype('float32')
inputs['scale_factor'] = np.array(
(im_info[0]['scale_factor'], )).astype('float32')
return inputs
for e in im_info:
im_shape.append(np.array((e['im_shape'], )).astype('float32'))
scale_factor.append(np.array((e['scale_factor'], )).astype('float32'))
inputs['im_shape'] = np.concatenate(im_shape, axis=0)
inputs['scale_factor'] = np.concatenate(scale_factor, axis=0)
imgs_shape = [[e.shape[1], e.shape[2]] for e in imgs]
max_shape_h = max([e[0] for e in imgs_shape])
max_shape_w = max([e[1] for e in imgs_shape])
padding_imgs = []
for img in imgs:
im_c, im_h, im_w = img.shape[:]
padding_im = np.zeros(
(im_c, max_shape_h, max_shape_w), dtype=np.float32)
padding_im[:, :im_h, :im_w] = img
padding_imgs.append(padding_im)
inputs['image'] = np.stack(padding_imgs, axis=0)
return inputs
class Detector(object):
def __init__(self,
pred_config,
model_dir):
self.pred_config = pred_config
self.predictor, self.config = load_predictor(model_dir)
self.preprocess_ops = self.get_ops()
def get_ops(self):
preprocess_ops = []
for op_info in self.pred_config.preprocess_infos:
new_op_info = op_info.copy()
op_type = new_op_info.pop('type')
preprocess_ops.append(eval(op_type)(**new_op_info))
return preprocess_ops
def predict(self, inputs):
input_names = self.predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.predictor.get_input_handle(input_names[i])
input_tensor.copy_from_cpu(inputs[input_names[i]])
self.predictor.run()
output_names = self.predictor.get_output_names()
boxes_tensor = self.predictor.get_output_handle(output_names[0])
np_boxes = boxes_tensor.copy_to_cpu()
boxes_num = self.predictor.get_output_handle(output_names[1])
np_boxes_num = boxes_num.copy_to_cpu()
results = []
if reduce(lambda x, y: x * y, np_boxes.shape) < 6:
results = {'boxes': np.zeros([]), 'boxes_num': [0]}
else:
results = {'boxes': np_boxes, 'boxes_num': np_boxes_num}
return results
def my_preprocess(para):
im_path, preprocess_ops = para
im, im_info = preprocess(im_path, preprocess_ops)
return im, im_info
def predict_image(detector, image_list, result_path):
c_results = {"result": []}
multiclass_thres = [0.49, 0.49, 0.49, 0.49, 0.49, 0.49, 0.49, 0.49]
num_worker = 4
pool = ThreadPool(processes=num_worker)
img_length = len(image_list)
img_iter_filter = 10
img_iter_range = list(range(img_length//img_iter_filter))
for start_index in img_iter_range:
if start_index == img_iter_range[-1]:
im_paths = image_list[start_index*img_iter_filter:]
else:
im_paths = image_list[start_index*img_iter_filter:(start_index+1)*img_iter_filter]
image_ids = [int(os.path.basename(im_p).split('.')[0]) for im_p in im_paths]
para = [[i,detector.preprocess_ops] for i in im_paths]
imandinfos = pool.map(my_preprocess, para)
for idx, imandinfo in enumerate(imandinfos):
image_id = image_ids[idx]
inputs = create_inputs([imandinfo[0]], [imandinfo[1]])
det_results = detector.predict(inputs)
im_bboxes_num = det_results['boxes_num'][0]
if im_bboxes_num > 0:
bbox_results = det_results['boxes'][0:im_bboxes_num, 2:]
id_results = det_results['boxes'][0:im_bboxes_num, 0]
score_results = det_results['boxes'][0:im_bboxes_num, 1]
for idx in range(im_bboxes_num):
if float(score_results[idx]) >= multiclass_thres[int(id_results[idx])]:
c_results["result"].append({"image_id": image_id,
"type": LABEL_MAP[str(int(id_results[idx]))],
"x": float(bbox_results[idx][0]),
"y": float(bbox_results[idx][1]),
"width": float(bbox_results[idx][2]) - float(bbox_results[idx][0]),
"height": float(bbox_results[idx][3]) - float(bbox_results[idx][1]),
"segmentation": []})
with open(result_path, 'w') as ft:
json.dump(c_results, ft)
def main(infer_txt, result_path, det_model_path):
pred_config = PredictConfig(det_model_path)
detector = Detector(pred_config, det_model_path)
img_list = get_test_images(infer_txt)
predict_image(detector, img_list, result_path)
if __name__ == '__main__':
print('start…')
start_time = time.time()
det_model_path = "model/"
paddle.enable_static()
infer_txt = sys.argv[1]
result_path = sys.argv[2]
main(infer_txt, result_path, det_model_path)
print('total time:', time.time() - start_time)
|
wechat
alipay
