-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdetect_1_resnet18.py
169 lines (156 loc) · 5.55 KB
/
detect_1_resnet18.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
# Author:yujunyu
# -*- codeing = utf-8 -*-
# @Time :2022/9/12 16:20
# @Author :yujunyu
# @Site :
# @File :test.py
# @software: PyCharm
import matplotlib.pyplot as plt
import torch
import numpy as np
from PIL import Image
from torchvision.transforms import Resize, Compose, ToTensor, Normalize, RandomHorizontalFlip, RandomVerticalFlip
from torchvision.models import resnet18
from torchvision.datasets import ImageFolder
import random
class GarbageRecognizer:
def __init__(self, module_file="./model/garbage_9.pth"):
super(GarbageRecognizer, self).__init__()
self.module_file = module_file
self.CUDA = torch.cuda.is_available()
self.net = resnet18(pretrained=False, num_classes=6)
if self.CUDA:
self.net.cuda()
device = 'cuda'
else:
device = 'cpu'
state = torch.load(self.module_file, map_location=device) # torch.load("./models", map_location=device)
self.net.load_state_dict(state)
print("加载模型完毕!")
self.net.eval()
@torch.no_grad()
def recognzie(self, img):
with torch.no_grad():
# 开始识别
if self.CUDA:
img = img.cuda()
# print(pre_img)
img = img.view(-1, 3, 224, 224)
y = self.net(img)
p_y = torch.nn.functional.softmax(y, dim=1)
p, cls_idx = torch.max(p_y, dim=1)
return cls_idx.cpu(), p.cpu()
# 识别测试
if __name__ == "__main__":
# 模型
model_file = 'model/resnet18_6.pth'
recognizer = GarbageRecognizer(model_file)
# # 下面转换用于独立的图像，并对其做预处理
transform = Compose(
[
Resize((224, 224)),
# RandomHorizontalFlip(), # 0.5的进行水平翻转
# RandomVerticalFlip(), # 0.5的进行垂直翻转
ToTensor(), # PIL转tensor
Normalize(mean=[0.6719, 0.6417, 0.6101], std=[0.2078, 0.2083, 0.2302])
]
)
dataset = ImageFolder('./dataset/image', transform=transform)
pre_dataset = ImageFolder("./input/pre_img1", transform=transform)
print(pre_dataset.class_to_idx)
'''
数据集随机预测
'''
sample_num = len(dataset)
print(sample_num)
pre_num = 10
samples_idx = random.sample(range(0, sample_num), pre_num)
print(f'数据集随机采样{pre_num}张')
# 循环识别
true_num1 = 0
label_list, pre_list, img_list = [], [], []
for idx in samples_idx:
cls, p = recognizer.recognzie(dataset[idx][0])
# print(dataset[idx][0]) # dataset[idx][0]像素 dataset[idx][1]类别标签（0~5）
real_cls1 = dataset.classes[dataset[idx][1]]
pre_cls1 = dataset.classes[cls]
label_list.append(real_cls1)
pre_list.append(pre_cls1)
img = dataset[idx][0]
img = img.swapaxes(0, 1)
img = img.swapaxes(1, 2)
img_list.append(img)
# plt.imshow(img)
# plt.show()
if real_cls1 == pre_cls1:
true_num1 += 1
print(
f'真实:{dataset[idx][1]}-{real_cls1} \t\t\t\t 预测:{cls.numpy()[0]}-{pre_cls1} \t\t\t\t 概率:{p.numpy()[0]}')
print(f'准确率:{true_num1 / pre_num}')
# print(label_list,pre_list)
fig = plt.gcf()
fig.set_size_inches(10, 12)
plt.rcParams['font.sans-serif'] = ['SimHei']
for i in range(0, pre_num):
ax = plt.subplot(4, 4, i + 1)
ax.imshow(img_list[i])
title = "true:" + str(label_list[i])
title += ", pre:" + str(pre_list[i])
ax.set_title(title, fontsize=10)
# ax.set_xticks([]) # 不显示坐标轴
# ax.set_yticks([])
plt.xlabel(f'准确率:{true_num1 / pre_num * 100}%', loc='center')
plt.show()
'''
预测图片随机预测
'''
# 样本总数
sample_num = len(pre_dataset)
print(sample_num)
# 随机采样，采样数 10
pre_num = 20
samples_idx = random.sample(range(0, sample_num), pre_num)
print(f'预测图片随机采样{pre_num}张')
# 循环识别
true_num = 0
label_list, pre_list, img_list = [], [], []
for idx in samples_idx:
cls, p = recognizer.recognzie(pre_dataset[idx][0])
# print(dataset[idx])
real_cls1 = pre_dataset.classes[pre_dataset[idx][1]]
pre_cls1 = pre_dataset.classes[cls]
label_list.append(real_cls1)
pre_list.append(pre_cls1)
img = dataset[idx][0]
img = img.swapaxes(0, 1)
img = img.swapaxes(1, 2)
img_list.append(img)
# 累加预测正确数量
if real_cls1 == pre_cls1:
true_num += 1
print(
f'真实:{pre_dataset[idx][1]}-{real_cls1} \t\t\t\t 预测:{cls.numpy()[0]}-{pre_cls1} \t\t\t\t 概率:{p.numpy()[0]}')
print(f'准确率:{true_num / pre_num}')
fig = plt.gcf()
fig.set_size_inches(10, 12)
for i in range(0, pre_num):
ax = plt.subplot(5, 4, i + 1)
ax.imshow(img_list[i])
title = "true:" + str(label_list[i])
title += ", pre:" + str(pre_list[i])
ax.set_title(title, fontsize=10)
ax.set_xticks([]) # 不显示坐标轴
ax.set_yticks([])
plt.xlabel(f'准确率:{true_num / pre_num * 100}%', loc='center')
plt.show()
'''
单张图片预测
'''
img_filename = './input/pre_img/metal/img_5.png'
print("预测单张图像：", img_filename)
img = Image.open(img_filename)
img = Image.fromarray(np.uint8(img))
img = transform(img)
cls, p = recognizer.recognzie(img)
cls = pre_dataset.classes[cls]
print(cls, '{:}%'.format(p.numpy()[0] * 100))