深度学习笔记-1.使用ANN分类animal实例

主要为了理解ANN使用方法

import cv2
import numpy as np
from random import randint

# ANN网络初始化
animals_net = cv2.ml.ANN_MLP_create()
animals_net.setTrainMethod(cv2.ml.ANN_MLP_RPROP | cv2.ml.ANN_MLP_UPDATE_WEIGHTS)
animals_net.setActivationFunction(cv2.ml.ANN_MLP_SIGMOID_SYM)
animals_net.setLayerSizes(np.array([3, 8, 4]))
animals_net.setTermCriteria(( cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 1 ))

"""Input arrays
weight, length, teeth
"""

"""Output arrays
dog, eagle, dolphin and dragon
"""

#狗，数据
def dog_sample():
  return [randint(10, 20), 1, randint(38, 42)]

#狗，标签
def dog_class():
  return [1, 0, 0, 0]

#秃鹰，数据
def condor_sample():
  return [randint(3,10), randint(3,5), 0]

#秃鹰，标签
def condor_class():
  return [0, 1, 0, 0]

#海豚，数据
def dolphin_sample():
  return [randint(30, 190), randint(5, 15), randint(80, 100)]

#海豚，标签
def dolphin_class():
  return [0, 0, 1, 0]

#龙，数据
def dragon_sample():
  return [randint(1200, 1800), randint(30, 40), randint(160, 180)]

#龙，标签
def dragon_class():
  return [0, 0, 0, 1]

#记录
def record(sample, classification):
  return (np.array([sample], dtype=np.float32), np.array([classification], dtype=np.float32))

records = []

"""
#不执行此段
SAMPLES = 5000
for x in range(0, SAMPLES):
  print ("Samples %d/%d" % (x, SAMPLES))
  animals_net.train(np.array([dog_sample()], dtype=np.float32), cv2.ml.ROW_SAMPLE, np.array([dog_class()], dtype=np.float32))
  animals_net.train(np.array([condor_sample()], dtype=np.float32), cv2.ml.ROW_SAMPLE, np.array([condor_class()], dtype=np.float32))
  animals_net.train(np.array([dolphin_sample()], dtype=np.float32), cv2.ml.ROW_SAMPLE, np.array([dolphin_class()], dtype=np.float32))
  animals_net.train(np.array([dragon_sample()], dtype=np.float32), cv2.ml.ROW_SAMPLE, np.array([dragon_class()], dtype=np.float32))
"""

#生成训练集
RECORDS = 5000
for x in range(0, RECORDS):
  records.append(record(dog_sample(), dog_class()))
  records.append(record(condor_sample(), condor_class()))
  records.append(record(dolphin_sample(), dolphin_class()))
  records.append(record(dragon_sample(), dragon_class()))


print("len records:%d"%len(records))


#迭代200轮，训练
EPOCHS = 2
for e in range(0, EPOCHS):
  print("Epoch %d:" % e)
  for t, c in records:
    animals_net.train(t, cv2.ml.ROW_SAMPLE, c)


#测试
TESTS = 100
dog_results = 0
for x in range(0, TESTS):
  clas = int(animals_net.predict(np.array([dog_sample()], dtype=np.float32))[0])
  print("class: %d" % clas)
  if (clas) == 0:
    dog_results += 1

condor_results = 0
for x in range(0, TESTS):
  clas = int(animals_net.predict(np.array([condor_sample()], dtype=np.float32))[0])
  print("class: %d" % clas)
  if (clas) == 1:
    condor_results += 1

dolphin_results = 0
for x in range(0, TESTS):
  clas = int(animals_net.predict(np.array([dolphin_sample()], dtype=np.float32))[0])
  print("class: %d" % clas)
  if (clas) == 2:
    dolphin_results += 1

dragon_results = 0
for x in range(0, TESTS):
  clas = int(animals_net.predict(np.array([dragon_sample()], dtype=np.float32))[0])
  print("class: %d" % clas)
  if (clas) == 3:
    dragon_results += 1

print ("Dog accuracy: %f%%" % (dog_results))
print ("condor accuracy: %f%%" % (condor_results))
print ("dolphin accuracy: %f%%" % (dolphin_results))
print ("dragon accuracy: %f%%" % (dragon_results))

原文地址：https://www.cnblogs.com/jsxyhelu/p/8611935.html

改进到自己的应用

import cv2
import numpy as np
import tensorflow as tf
import random

from random import randint



#字符图像的特征提取方法
#要求：1.无关图像大小；2.输入图像默认为灰度图
def SimpleGridFeature(image):
    '''
    @description:提取字符图像的简单网格特征
    @image:灰度字符图像
    @return:长度为64字符图像的特征向量feature
    @author:RenHui
    '''
    #图像扩充
    resize_flag = 0
    if image.shape[0]%8 !=0:
        img_h = image.shape[0] + (8 - image.shape[0]%8)
        resize_flag = 1
    if image.shape[1]%8 !=0:
        img_w = image.shape[1] + (8 - image.shape[1]%8)
        resize_flag = 1
    if resize_flag:
        image = cv2.resize(image,(img_w,img_h))

    #二值化
    retval,binary = cv2.threshold(image,0,255,cv2.THRESH_OTSU)
    #计算网格大小
    grid_h = binary.shape[0]/8
    grid_w = binary.shape[1]/8
    #定义特征向量
    feature = np.zeros(64)
    for j in range(8):
        for i in range(8):
            grid = binary[int(j*grid_h):int((j+1)*grid_h),int(i*grid_w):int((i+1)*grid_w)]
            feature[j*8+i] = grid[grid==0].size
    return feature

#OCR，类标签
# i范围 0-9
def OCR_classLabel(i):
    list = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
    list[i]=1
    return list

#记录
def record(sample, classification):
  return (np.array([sample], dtype=np.float32), np.array([classification], dtype=np.float32))

#产生随机数
#random_int_list(1, 1, 64)
def random_int_list(start, stop, length):
  start, stop = (int(start), int(stop)) if start <= stop else (int(stop), int(start))
  length = int(abs(length)) if length else 0
  random_list = []
  for i in range(length):
    random_list.append(random.randint(start, stop))
  return random_list

############################################################################



# ANN网络初始化
OCR_ANN = cv2.ml.ANN_MLP_create()    #创建
OCR_ANN.setTrainMethod(cv2.ml.ANN_MLP_RPROP | cv2.ml.ANN_MLP_UPDATE_WEIGHTS)
OCR_ANN.setActivationFunction(cv2.ml.ANN_MLP_SIGMOID_SYM)   #激活函数
OCR_ANN.setLayerSizes(np.array([64, 64, 10]))    #每层神经元
OCR_ANN.setTermCriteria(( cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 1 ))


#生成训练集
traindata = []
features = np.load('features_0-9.npy')
labels = np.load('labels_0-9.npy')
for feature,label in zip (features,labels):
    traindata.append(record(feature,label))
# RECORDS = 50
# for x in range(0, RECORDS):
#     if (x % 1000 == 0):
#         print("生成数据集：%d"% x)
#     traindata.append(record(random_int_list(0, 0, 64),OCR_classLabel(0)))
#     traindata.append(record(random_int_list(1, 1, 64),OCR_classLabel(1)))
#     traindata.append(record(random_int_list(2, 2, 64),OCR_classLabel(2)))
#     traindata.append(record(random_int_list(3, 3, 64),OCR_classLabel(3)))
#     traindata.append(record(random_int_list(4, 4, 64),OCR_classLabel(4)))
#     traindata.append(record(random_int_list(5, 5, 64),OCR_classLabel(5)))
#     traindata.append(record(random_int_list(6, 6, 64),OCR_classLabel(6)))
#     traindata.append(record(random_int_list(7, 7, 64),OCR_classLabel(7)))
#     traindata.append(record(random_int_list(8, 8, 64),OCR_classLabel(8)))
#     traindata.append(record(random_int_list(9, 9, 64),OCR_classLabel(9)))


#打乱训练数据集
print("随机打乱数据中...")
random.shuffle(traindata)  #打乱数

# 迭代200轮，训练
EPOCHS = 200
for e in range(0, EPOCHS):
    print("Epoch %d:" % e)
    counter = 0
    for data, label in traindata:
        # print("data:")
        # print(data)
        # print("label:")
        # print(label)
        counter +=1
        if (counter % 1000 == 0):
            print("Epoch %d: Trained %d" % (e, counter))
        OCR_ANN.train(data, cv2.ml.ROW_SAMPLE, label)


#测试


# 读图
path = "E:/sxl_Programs/Python/ANN/2.jpg"
image = cv2.imread(path,0)
#获取特征向量，传入灰度图
ArrayFeature=SimpleGridFeature(image)
print("ArrayFeature:",ArrayFeature)
clas = int(OCR_ANN.predict(np.array([ArrayFeature], dtype=np.float32))[0])
print("class: %d" % clas)

# TESTS = 10
# dog_results = 0
# for x in range(0, TESTS):
#   clas = int(OCR_ANN.predict(np.array([random_int_list(1,1, 64)], dtype=np.float32))[0])
#   print("class: %d" % clas)

欢迎扫码关注我的微信公众号