-
Notifications
You must be signed in to change notification settings - Fork 3
/
prediction.py
133 lines (106 loc) · 3.55 KB
/
prediction.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
# -*- coding: utf-8 -*-
# @Time : May 25
# @Author : Xuyang SHEN, Alisdair Cameron, Xinqi Zhu
# @File : prediction.py
# @IDE: PyCharm Community Edition
import tensorflow as tf
import os
import matplotlib.pyplot as plt
import cv2
import numpy as np
import imageio
from enet.ENet import *
from config import *
from data_provider.one_hot_coding import one_hot_decode
def transparent_circle(img, center, radius, rgb_alpha):
center = tuple(map(int, center))
rgb = rgb_alpha[:3]
alpha = rgb_alpha[-1]
radius = int(radius)
roi = slice(center[1] - 3, center[1] + 3), slice(center[0] - 3, center[0] + 3)
overlay = img[roi].copy()
cv2.circle(img, center, radius, rgb, thickness=0, lineType=cv2.LINE_AA)
cv2.addWeighted(src1=img[roi], alpha=alpha, src2=overlay, beta=1. - alpha, gamma=0, dst=img[roi])
def display_plot(origin, one_hot, comb):
fig = plt.figure(figsize=(10, 33))
ax1 = fig.add_subplot(311)
origin = np.reshape(origin, (512, 512, 3))
origin = origin.astype('uint8')
ax1.imshow(origin)
ax1.set_title("origin image")
ax2 = fig.add_subplot(312)
ax2.imshow(one_hot, cmap='Greys')
ax2.set_title("one_hot image")
ax2 = fig.add_subplot(313)
ax2.imshow(comb)
ax2.set_title("combine the one_hot into original image")
plt.show()
# ---------------------------------------------------------
# run parser
# ---------------------------------------------------------
config = parse_cmd_testing_args()
test_add = config.test_dataset
model_add = config.model
output_add = config.result_address
display = config.display
# ---------------------------------------------------------
# initialized the model
# ---------------------------------------------------------
model_address = os.path.join(os.getcwd(), model_add)
lane_detect = tf.estimator.Estimator(
model_fn=ENet,
model_dir=model_address,
)
# ---------------------------------------------------------
# prediction data set
# ---------------------------------------------------------
pred_images = []
filename = os.listdir(test_add)
for f in filename:
im = plt.imread(os.path.join(test_add, f))
eval_data = cv2.resize(im, (512, 512))
eval_data = eval_data.astype('float32')
eval_data = np.reshape(eval_data, (1, 512, 512, 3))
pred_images.append(eval_data)
pred_images = np.array(pred_images)
# ---------------------------------------------------------
# prediction the model
# ---------------------------------------------------------
pred_input_fn = tf.estimator.inputs.numpy_input_fn(
x=pred_images,
shuffle=False,
batch_size=1
)
pred_result = lane_detect.predict(
input_fn=pred_input_fn,
checkpoint_path=None
)
# iteration all prediction
var = 0
for i in pred_result:
i = i['classes']
re = np.zeros(shape=(512, 512))
for col in range(512):
for row in range(512):
# focus on the probability
if i[col, row, 0] < i[col, row, 1]:
re[col, row] = 255
# one_hot labels
re = np.reshape(re, (512, 512))
re = re.astype('uint8')
path = output_add + "pred-" + filename[var]
imageio.imwrite(path, re)
# combine original with current
coords = one_hot_decode(re)
im = pred_images[var]
im = np.reshape(im, (512, 512, 3))
im = im.astype('uint8')
for coord in coords:
transparent_circle(img=im, center=coord, radius=1, rgb_alpha=[102, 255, 51, 0.2])
path = output_add + "comb-" + filename[var]
imageio.imwrite(path, im)
if display == 'yes':
display_plot(
pred_images[var], re, im
)
var += 1