-
-
一个图片分类器_神经网络_反向传播
-
发表于: 2022-9-18 11:14
-
说明
1 | 一个练习代码，使用神经网络分类10张数字图片。 |
代码
main.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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 | from PIL import Imageimport numpy as npimport sysfrom datetime import datetimeimport jsonclass Dot(object): def forward(self, X, W): self.X=X self.W=W return np.dot(X, W) def backward(self, dout): return np.dot(dout, self.W.T), np.dot(self.X.T, dout) # dX, dWclass Add(object): def forward(self, D, B): return D+B def backward(self, dout): return np.sum(dout, axis=0) #纵向求和class Affine(object): def __init__(self): self.dot=Dot() self.add=Add() def forward(self, X, W, B): T=self.dot.forward(X, W) return self.add.forward(T, B) def backward(self, dout): dB = self.add.backward(dout) dX, dW = self.dot.backward(dout) return dX, dW, dBclass Relu(object): def forward(self, A): self.mask=(A<=0) A[self.mask]=0 #会破坏原参数，但是这次使用之后后面就没用了，所以破坏就破坏了，还能省点空间 return A def backward(self, dout): dout[self.mask]=0 return doutclass SoftmaxWithLoss(object): def softmax(self, A): A=A-A.max(axis=1, keepdims=True) T=np.exp(A) return T/np.sum(T, axis=1, keepdims=True) def crossEntropyError(self, Z, Label): delta=0.000000001 return -np.sum(np.log(Z+delta)*Label)/Z.shape[0] def forward(self, A, Label): self.Z=self.softmax(A) self.Label=Label return self.Z, self.crossEntropyError(self.Z, Label) def backward(self, dout=1): return (self.Z-self.Label)*dout/self.Z.shape[0]class SimpleImageClassifier(object): def __init__(self): # 初始化一个3层网络 # 学习率设置为 0.01 self.N0=28*28 self.N1=10 self.N2=10 self.N3=10 self.lr=0.01 self.MAX_NUM=100000 self.affine1=Affine() self.relu1 =Relu() self.affine2=Affine() self.relu2 =Relu() self.affine3=Affine() self.softmax_with_loss=SoftmaxWithLoss() ''' self.W1=np.random.randn(self.N0, self.N1) self.B1=np.random.randn(self.N1) self.W2=np.random.randn(self.N1, self.N2) self.B2=np.random.randn(self.N2) self.W3=np.random.randn(self.N2, self.N3) self.B3=np.random.randn(self.N3) ''' self.loadWB() print("self.W1\n",self.W1) print("self.B1\n",self.B1) print("self.W2\n",self.W2) print("self.B2\n",self.B2) print("self.W3\n",self.W3) print("self.B3\n",self.B3) print("") def loadImg(self): lst_img = [] for i in range(0, 10): img = Image.open("img/{num}.jpg".format(num=i)) lst_img.append({"data":img.tobytes(), "label":i}) return lst_img def predictWithLoss(self, X, Label): # 0 [1, 0, 0, 0, 0, 0, 0, 0, 0, 0] -> [0.8, 0, 0.1, 0, 0.05, 0.05, 0, 0, 0, 0] # 1 [0, 1, 0, 0, 0, 0, 0, 0, 0, 0] # 2 [0, 0, 1, 0, 0, 0, 0, 0, 0, 0] ... # 输入是 n*784的向量，n张图片，每张图有784个值 (784=28*28) A1=self.affine1.forward(X, self.W1, self.B1) Z1= self.relu1.forward(A1) A2=self.affine2.forward(Z1, self.W2, self.B2) Z2= self.relu2.forward(A2) A3=self.affine3.forward(Z2, self.W3, self.B3) Z3, l = self.softmax_with_loss.forward(A3, Label) return Z3, l def getNumericalGradient(self): # 使用反向传播计算W与B的偏导数 D = self.softmax_with_loss.backward(dout=1) dZ2, dW3, dB3 = self.affine3.backward(D) dA2 = self.relu2.backward(dZ2) dZ1, dW2, dB2 = self.affine2.backward(dA2) dA1 = self.relu1.backward(dZ1) dX, dW1, dB1 = self.affine1.backward(dA1) ''' print("grad_W1\n", dW1) print("grad_B1\n", dB1) print("grad_W2\n", dW2) print("grad_B2\n", dB2) print("grad_W3\n", dW3) print("grad_B3\n", dB3) ''' return dW1,dB1,dW2,dB2,dW3,dB3 def updateWB(self, grad_W1,grad_B1,grad_W2,grad_B2,grad_W3,grad_B3): def updateW(W, grad_W): W-=self.lr*grad_W def updateB(B, grad_B): B-=self.lr*grad_B updateW(self.W1, grad_W1) updateB(self.B1, grad_B1) updateW(self.W2, grad_W2) updateB(self.B2, grad_B2) updateW(self.W3, grad_W3) updateB(self.B3, grad_B3) def train(self): # 输入数据初始化 X=[] Label=[] lst_img=self.loadImg() for dic in lst_img: b=max(dic["data"][:self.N0]) X.append([ a/b for a in dic["data"][:self.N0] ]) # 归一化 # t=[0]*self.N3 t[dic["label"]]=1 Label.append(t) X=np.array( X ) Label=np.array( Label ) self.showDetail(X, Label, i=-1) # 算偏微分，更新权重参数 for i in range(0, self.MAX_NUM): self.predictWithLoss(X, Label) grad_W1,grad_B1,grad_W2,grad_B2,grad_W3,grad_B3 = self.getNumericalGradient() self.updateWB(grad_W1,grad_B1,grad_W2,grad_B2,grad_W3,grad_B3) ''' if i%20==0: print(i) print("self.W1\n",self.W1) print("self.B1\n",self.B1) print("self.W2\n",self.W2) print("self.B2\n",self.B2) print("self.W3\n",self.W3) print("self.B3\n",self.B3) print("") ''' if i%20==0: self.showDetail(X, Label, i) self.showDetail(X, Label, i=self.MAX_NUM) def showDetail(self, X, Label, i=0): msg="i={i} time:{time}".format(i=i, time=datetime.now().strftime("%Y-%m-%d %H:%M:%S")) print(msg) Z, l = self.predictWithLoss(X, Label) print(Z) print(l) print("") self.saveWB(out_file="wb.json", additional_data={"i":msg, "Z":Z.tolist(), "loss":l}) def saveWB(self, out_file="wb.json", additional_data={}): dic={"W1":self.W1.tolist(), "B1":self.B1.tolist(), "W2":self.W2.tolist(), "B2":self.B2.tolist(), "W3":self.W3.tolist(), "B3":self.B3.tolist(), "additional_data":additional_data} fout=open(out_file, "a+") fout.write(json.dumps(dic)) fout.write("\n") fout.close() def loadWB(self, in_file="wb.json"): WB='''{"W1": [[0.7189550494057414, 0.1989326646832533,...示意..., "loss": 0.008417616242222797}}''' dic=json.loads(WB) self.W1=np.array( dic["W1"] ) self.B1=np.array( dic["B1"] ) self.W2=np.array( dic["W2"] ) self.B2=np.array( dic["B2"] ) self.W3=np.array( dic["W3"] ) self.B3=np.array( dic["B3"] ) def whichNumber(self, img_file): img = Image.open(img_file) data=img.tobytes() X=[] b=max(data[:self.N0]) X.append([ a/b for a in data[:self.N0] ]) Z, l=self.predictWithLoss(X, [0]*10) print(Z[0]) b=max(Z[0]) num=np.where(Z[0]==b)[0][0] print("num={num}, rate={rate}".format(num=num, rate=b)) return numdef main(): a=SimpleImageClassifier() #a.train() a.whichNumber("img/9.jpg")if "__main__"==__name__: main() |
运行截图
参考
《深度学习入门：基于Python的理论与实现》
[招生]科锐逆向工程师培训(2024年11月15日实地，远程教学同时开班, 第51期)
最后于 2022-10-4 16:15
被Jtian编辑
，原因：
