lenet_test.zip_ai_keras_lenet_lenet test_tensorflow

# -*- coding: utf-8 -*- #Author: weir #E-mail: weirmicle@163.com #Date: 2017/12/23 #the kernel model is from http://blog.csdn.net/wang1127248268/article/details/77258055 import numpy as np import matplotlib.pyplot as plt import cv2 import struct import tensorflow as tf import keras.backend.tensorflow_backend as ktf import keras as ks from keras.utils import np_utils from keras.models import Sequential from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D, Activation from keras.layers.convolutional import Conv2D,MaxPooling2D from keras.utils.np_utils import to_categorical ###################### input_imgs_num = 20000#max 60_000 test_imgs_num = 10000#max 10_000 batch_size = 50 epochs=20 ###################### def get_session(gpu_fraction=0.333): gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=gpu_fraction, allow_growth=True) return tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) # train path: train_images_idx3_ubyte_file = './data/train-images-idx3-ubyte' train_labels_idx1_ubyte_file = './data/train-labels-idx1-ubyte' # test path: test_images_idx3_ubyte_file = './data/t10k-images-idx3-ubyte' test_labels_idx1_ubyte_file = './data/t10k-labels-idx1-ubyte' def decode_idx3_ubyte_img(idx3_ubyte_file): """ :param idx3_ubyte_file: the path of file :return ndarray byte[0 :3 ]: magic number, the file formatter flag byte[4 :7 ]: id0,img_batches byte[8 :11]: id1,img_rows byte[12:15]: id2,img_cols byte[16: ]: pixels of ubyte """ binfile = open(idx3_ubyte_file, "rb") bin_head = binfile.read(16) magic_num, img_batches, img_rows, img_cols = struct.unpack(">4i", bin_head) bin_data = binfile.read() fmt = ">%dB" %(img_batches*img_rows*img_cols) imgs_ld = struct.unpack(fmt, bin_data) imgs_2d = np.array(imgs_ld).reshape([img_batches, img_rows, img_cols]) imgs_2d = imgs_2d.astype(np.uint8) return imgs_2d def decode_idx1_ubyte_img(idx1_ubyte_file): """ :param idx3_ubyte_file: the path of file :return ndarray byte[0 :3 ]: magic number, the file formatter flag byte[4 :7 ]: id0,number of items """ binfile = open(idx1_ubyte_file, "rb") bin_head = binfile.read(8) magic_num, num_items = struct.unpack(">2i", bin_head) bin_data = binfile.read() fmt = ">%dB" %(num_items) imgs_ld = struct.unpack(fmt, bin_data) imgs_ld = np.array(imgs_ld, dtype=np.uint8) return imgs_ld def img_3dto4d(img): print("in 3dto4d(),img.shape", img.shape,"img.dtype: ", img.dtype) return img.reshape(img.shape[0], 28, 28, 1) def main(): ktf.set_session(get_session()) #get train set: train_imgs = decode_idx3_ubyte_img(train_images_idx3_ubyte_file) train_label = decode_idx1_ubyte_img(train_labels_idx1_ubyte_file) #get test set: test_imgs = decode_idx3_ubyte_img(test_images_idx3_ubyte_file) test_label = decode_idx1_ubyte_img(test_labels_idx1_ubyte_file) train_img4d = img_3dto4d(train_imgs) test_img4d = img_3dto4d(test_imgs) labels_onehot = np_utils.to_categorical(train_label, num_classes=10)#to onehot test_labels_onehot = np_utils.to_categorical(test_label, num_classes=10)#to onehot x_imgs = train_img4d[0:input_imgs_num] y_labels = labels_onehot[0:input_imgs_num] x_imgs.astype('float32') x_imgs = x_imgs/255; #model = ks.models.Sequential() #model.add(Conv2D(filters=128, kernel_size=(3,3), strides=1, input_shape=(28,28,1),padding='valid', activation='relu')) #model.add(Conv2D(32,kernel_size=(3,3),activation='relu')) #model.add(MaxPooling2D(pool_size=(2, 2))) #model.add(Dropout(0.25)) #model.add(Flatten()) #model.add(Dense(128, activation='relu')) #model.add(Dropout(0.5)) #model.add(Dense(10, activation='softmax')) model = Sequential() model.add(Conv2D(32,(5,5),strides=(1,1),input_shape=(28,28,1),padding='valid',activation='relu',kernel_initializer='uniform')) model.add(MaxPooling2D(pool_size=(2,2))) model.add(Conv2D(64,(5,5),strides=(1,1),padding='valid',activation='relu',kernel_initializer='uniform')) model.add(MaxPooling2D(pool_size=(2,2))) model.add(Flatten()) model.add(Dense(100,activation='relu')) model.add(Dense(10,activation='softmax')) model.compile(optimizer='sgd',loss='categorical_crossentropy',metrics=['accuracy']) model.summary() #model.compile(loss='categorical_crossentropy',optimizer='adadelta',metrics=['accuracy']) model.fit(x=x_imgs,y=y_labels, batch_size=batch_size,epochs=epochs,verbose=2) y=model.predict(test_img4d[0:test_imgs_num]) #print("the predict result output is",y) res=np.array([np.argmax(i) for i in y]) #print("---the real digits is", test_label[0:test_imgs_num]) #print("the predict digits is",res) match = 0 for i in range(test_imgs_num): if test_label[i] == res[i]: match = match + 1 match_ratio = 1.0 * match / test_imgs_num print(match, "/", test_imgs_num, "match_ratio = ", match_ratio) eval = model.evaluate(test_img4d, test_labels_onehot, 128) print(eval) if __name__ == "__main__": main()