图片转换为batch批数据（25）---《深度学习》

将图片转换为tensorflow中可以使用的批数据

#coding=utf-8
from PIL import Image
import os
import os.path
import numpy as np
import tensorflow as tf
import tensorflow.contrib.slim as slim
import tensorflow.contrib.slim.nets as nets
import inception_resnet_v2

def convert(dir):
    filelists=os.listdir(dir)
    arr_col=[]
    for file in filelists:
        file_path=os.path.join(dir,file)
        img=Image.open(file_path).resize((299,299)).convert("RGB")
        r,g,b=img.split()

        r_arr=np.array(r)
        g_arr=np.array(g)
        b_arr=np.array(b)

        img_arr=np.concatenate((r_arr,g_arr,b_arr))
        result=img_arr.reshape((299,299,3))
        arr_col.append(result)
    return arr_col
def convert_3_2_4_dims(arr_):
    ret=np.zeros((len(arr_),arr_[0].shape[0],arr_[0].shape[1],arr_[0].shape[2]))
    for i in range(len(arr_)):
        ret[i,:,:,:]=arr_[i]
    return ret
def to_categorial(y,n_classes):
    y_std=np.zeros([len(y),n_classes])
    for i in range(len(y)):
        y_std[i,y[i]]=1.0
    return y_std
#注意这儿进行选用list而不是tuple的原因在于后面数据的归一化需要改变值，而tuple不支持这种操作，所以选用list
def batch_list(x,y,batch_size):
    batches=[]
    for i in range(int(len(x)/batch_size)):
        print("enen"+str(i))
        batch_data=[x[batch_size*i:batch_size*i+batch_size],y[batch_size*i:batch_size*i+batch_size]]
        batches.append(list(batch_data))
    if(i+1)*batch_size<len(x):
        batch_data=[x[batch_size*(i+1):],y[batch_size*(i+1):]]
        batches.append(list(batch_data))

    return batches
def load_data(dir,num_classes,batch_size):
    arr_col=convert_3_2_4_dims(convert(dir))
    arr_col=arr_col.astype(np.float32)
    #因为这儿我没指定它的标签，所以就随机指定了一些标签
    z=np.random.rand(arr_col.shape[0])*num_classes
    z=z.astype("int")
    labels=np.array(z)
    batches=batch_list(arr_col,to_categorial(labels,num_classes),batch_size)
    return batches
#对数据进行归一化
def data_lrn(batches):
    for batch in batches:
        print(type(batch[0]))
        batch[0]/=255
    return batches