资源简介
vgg_easy.py
代码片段和文件信息
import os
import numpy as np
import tensorflow as tf
from PIL import Image
from skimage import io transform
import glob
os.environ[“CUDA_VISIBLE_DEVICES“] = “0“
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.9)
path = ‘./easy/‘ # for one dataset cross validation
train_path = ‘./example/train/‘ # for train and test set
test_path = ‘./example/test/‘
w = 224
h = 224
c = 3
n_class = 4
def read_img(path):
cate = [path + x for x in os.listdir(path) if os.path.isdir(path + x)]
imgs = []
labels = []
label_list = np.eye(n_class)
for idx folder in enumerate(cate): #search folder
for im in glob.glob(folder + ‘/*.jpg‘): #change doc type if necessary
img = io.imread(im)
img = transform.resize(img (w h c))
imgs.append(img) # (sum2242243)
labels.append(label_list[idx]) # (sum4)
return np.asarray(imgs np.float32) np.asarray(labels np.float32)
#------------------train and test set------------
data label = read_img(train_path)
num_example = data.shape[0]
arr = np.arange(num_example)
np.random.shuffle(arr) # random sequence
x_train = data[arr]
y_train = label[arr]
s = num_example
data_t label_t = read_img(test_path)
s_test = data_t.shape[0]
arr = np.arange(s_test)
np.random.shuffle(arr) # random sequence
x_val = data_t[arr]
y_val = label_t[arr]
# ------------------one dataset cross validation ----------
#data label = read_img(path)
#num_example = data.shape[0]
#arr = np.arange(num_example)
#np.random.shuffle(arr) # random sequence
#data = data[arr]
#label = label[arr]
#ratio = 0.8
#s = np.int(num_example * ratio)
#x_train = data[:s] # (sum_train2242243)
#y_train = label[:s] # (sum_train4)
#x_val = data[s:]
#y_val = label[s:]
#------------------vgg16 structure----------------
x = tf.placeholder(tf.float32 shape=[None h w c])
y = tf.placeholder(tf.float32 shape=[None n_class])
#----------------- conv1 ------------------------
w_conv1_1 = tf.Variable(tf.truncated_normal([3 3 3 64] stddev=0.1))
b_conv1_1 = tf.Variable(tf.constant(0.1 shape=[64]))
L_conv1_1 = tf.nn.relu(tf.nn.conv2d(x w_conv1_1strides=[1 1 1 1] padding=‘SAME‘) + b_conv1_1)
w_conv1_2 = tf.Variable(tf.truncated_normal([3 3 64 64] stddev=0.1))
b_conv1_2 = tf.Variable(tf.constant(0.1 shape=[64]))
L_conv1_2 = tf.nn.relu(tf.nn.conv2d(L_conv1_1 w_conv1_2strides=[1 1 1 1] padding=‘SAME‘) + b_conv1_2)
L_pool1 = tf.nn.max_pool(L_conv1_2 ksize=[1 2 2 1] strides=[1 2 2 1] padding=‘SAME‘)
#----------------- conv2 ------------------------
w_conv2_1 = tf.Variable(tf.truncated_normal([3 3 64 128
- 上一篇:keywordsearch
- 下一篇:基于多变量线性回归的房屋销售价格预测.zip
相关资源
- scel2txt.py
- 小甲鱼Python零基础免费全套视频教学
- 传智播客python最新视频.txt
- Python升级3.6强力Django杀手级Xadmin打造
- shuake.py
- 51Job.py
- python3网络爬虫与开发实战崔庆才PDF百
- python300G视频书籍教程.zip
- 5652华尔街见闻.py
- hulk.py
- 麦子学院Python视频.txt
- python核心基础.txt
- 仿真3D版本.py
- Python零基础10天进阶班.rar
- 用Python自动办公,做职场高手.txt
- 基于python实现的http接口自动化测试框
- spider_LOL.py
- Python教程.rar
- 小甲鱼零基础入门学习Python视频教程
- 麦子学院Python全套视频.txt
- 《疯狂Python讲义》习题答案.rar
- publishHelper.py
- sendInfo.py
- Python零基础10天进阶班.docx
- Python数据科学指南_Code.zip
- 爬取URP教务网站学籍信息.py
- Anaconda历史版本Python3.6版本.zip
- car_detected.py
- Python-图像分割Keras在Keras中实现Segne
评论
共有 条评论