准备工作
需要把实例图片信息(文件名)按照大图顺序排列,保证一个大图切出来的实例是紧挨着的,并且文件名有判断价值,因为后续主要用文件名判断是否属于同一个大图。
一、找到网络模型的预测网络入口
from classification import Classification
classfication = Classification()
二、读取数据信息,设置初始化变量
代码如下(示例):
f = open('.../***.txt', 'r')
files = f.readlines() # 读取整个文件所有行,保存在 list 列表中
[la,lb] = [0,0]
tmp = '0'
correct_num=0
error_num=0
[a,b] = [1,1]
三、循环判断
代码如下(示例):
for filename in files:
filename = filename[:-1]
image = Image.open('image_path'+filename)
class_name, probability= classfication.detect_image(image)
if filename[:12] == tmp:
if class_name == 'A' :
la = la+1
elif class_name=='B':
lb = lb+1
elif filename[:12] != tmp:
if class_name == 'A':
la = 1
lb = 0
elif class_name == 'B':
la = 0
lb = 1
m = max(a,b)
if m==a:
print(tmp+',0,正确')
# print(tmp+'预测成功,acc为:{0}'.format(acc))
correct_num = correct_num+1
elif m==b:
print(tmp + ',1,失败')
# print(tmp + '预测失败')
error_num = error_num+1
tmp = filename[:12]
a = la
b = lb