改进CNN代码注解

可能看着不是那么明确，但是每一句，每个方法都查过之后还是收获满满。fighting------
def OHE_labels(Y_tr,N_classes):（一）功能：返回Y_labels这个数组
    OHC = OneHotEncoder()
    Y_ohc = OHC.fit(np.arange(N_classes).reshape(-1, 1))
    ##arange将这么些类生成一维数组，reshape将其转化为多维，此处将其重塑成？行，1列。-1表示待定
    Y_labels = Y_ohc.transform(Y_tr.reshape(-1, 1)).toarray()
    return Y_labels

training_file = 'TrafficSignData/train.p'
testing_file = 'TrafficSignData/test.p'
with open(training_file, mode='rb') as f:
    train = pickle.load(f)
with open(testing_file, mode='rb') as f:
    test = pickle.load(f)   
X_train, y_train = train['features'], train['labels']
##X_train, y_train得到的是字典的键
X_test, y_test = test['features'], test['labels']
labels_train = OHE_labels(y_train,43)
##y_train的labels经过函数一变为43行1列的矩阵再变为数组：相当于转置吧！
labels_test = OHE_labels(y_test,43)

def check_OHE(cls,y):（二）功能：核实函数一的功能
    check = np.linalg.norm(np.argmax(cls,axis=1)-y)
    if check == 0:
        print('One hot encoding correct, ok to proceed')
    else:
        print('One hot encoding doesnt match the output, check code!!!')
##检查上一步的one-hot码是否正确

n_train = len(y_train)
n_test = len(y_test)
image_shape = np.shape(X_train[1])
##数据集【1】那里是数字的类吧！！！
n_classes = len(np.unique(y_train))
##len:取出来唯一那个数放在容器里返回容器中的项数。

def plot_random_3C(n_row,n_col,X,y):（三）功能

    plt.figure(figsize = (11,8))
    ##画一个图，宽11，高8
    gs1 = gridspec.GridSpec(n_row,n_row)
    ##划分成多少行多少列
    gs1.update(wspace=0.01, hspace=0.02) 
    # 设置轴之间的空隙 
    for i in range(n_row*n_col):
        ##  i = i + 1 下标从0开始
        ax1 = plt.subplot(gs1[i])
        plt.axis('on')
        ax1.set_xticklabels([])
        ax1.set_yticklabels([])
        ax1.set_aspect('equal')
        ind_plot = np.random.randint(1,len(y))
        ##随机得到长度为y，最小为1的数组
        plt.imshow(X[ind_plot])
        plt.text(2,4,str(y[ind_plot]),
             color='k',backgroundcolor='c')
        plt.axis('off')
    plt.show()
def plot_random_1C(n_row,n_col,X,y):

    plt.figure(figsize = (11,8))
    gs1 = gridspec.GridSpec(n_row,n_row)
    gs1.update(wspace=0.01, hspace=0.02) # set the spacing between axes. 

    for i in range(n_row*n_col):
        # i = i + 1 # grid spec indexes from 0
        ax1 = plt.subplot(gs1[i])
        plt.axis('on')
        ax1.set_xticklabels([])
        ax1.set_yticklabels([])
        ax1.set_aspect('equal')
        #plt.subplot(4,11,i+1)
        ind_plot = np.random.randint(1,len(y))
        plt.imshow(X[ind_plot],cmap='gray')
        plt.text(2,4,str(y[ind_plot]),
             color='k',backgroundcolor='c')
        plt.axis('off')
    plt.show()   
def plot_random_preprocess(n_row,n_col,X,y):

    plt.figure(figsize = (11,8))
    gs1 = gridspec.GridSpec(n_row,n_row)
    gs1.update(wspace=0.01, hspace=0.02) # set the spacing between axes. 

    for i in range(n_row*n_col):
        # i = i + 1 # grid spec indexes from 0
        ax1 = plt.subplot(gs1[i])
        plt.axis('on')
        ax1.set_xticklabels([])
        ax1.set_yticklabels([])
        ax1.set_aspect('equal')
        #plt.subplot(4,11,i+1)
        ind_plot = np.random.randint(1,len(y))
        plt.imshow(pre_process_image(X[ind_plot]),cmap='gray')
        plt.text(2,4,str(y[ind_plot]),
             color='k',backgroundcolor='c')
        plt.axis('off')
    plt.show()

    
def pre_process_image(image):

    #image = cv2.cvtColor(image, cv2.COLOR_BGR2YUV)只保留明亮度（Y）色度（U、V）
    #image = image[:,:,0]
    image[:,:,0] = cv2.equalizeHist(image[:,:,0])
    image[:,:,1] = cv2.equalizeHist(image[:,:,1])
    image[:,:,2] = cv2.equalizeHist(image[:,:,2])
    ##直方图均衡化
    image = image/255.-.5
    #image = cv2.resize(image, (img_resize,img_resize),interpolation = cv2.INTER_CUBIC)
    
    return image
.......中间有个问题：

plt.figure(figsize=(12,8))
plt.bar(range(43),height=data_pd_sorted["Occurance"])

在anaconda里可以实现，在pycharm实现不了，既不可以放在main函数里，也不能用print，改为下面的语句即可实现

height = data_pd_sorted["Occurance"]#先把要画的图定成一个常量
plt.figure(figsize=(12,8))#进行坐标轴建立
p1=plt.bar(range(43),height=data_pd_sorted["Occurance"])#用不用p1变量承接都可以
plt.show()#show才是可以让这个图画出来的关键方法

 

img_resize = 32
def transform_image(image,ang_range,shear_range,trans_range):
    ang_rot = np.random.uniform(ang_range)-ang_range/2
    rows,cols,ch = image.shape    
    Rot_M = cv2.getRotationMatrix2D((cols/2,rows/2),ang_rot,1)

    # Translation
    tr_x = trans_range*np.random.uniform()-trans_range/2
    tr_y = trans_range*np.random.uniform()-trans_range/2
    Trans_M = np.float32([[1,0,tr_x],[0,1,tr_y]])

    # Shear
    pts1 = np.float32([[5,5],[20,5],[5,20]])

    pt1 = 5+shear_range*np.random.uniform()-shear_range/2
    pt2 = 20+shear_range*np.random.uniform()-shear_range/2

    pts2 = np.float32([[pt1,5],[pt2,pt1],[5,pt2]])

    shear_M = cv2.getAffineTransform(pts1,pts2)
        
    image = cv2.warpAffine(image,Rot_M,(cols,rows))
    image = cv2.warpAffine(image,Trans_M,(cols,rows))
    image = cv2.warpAffine(image,shear_M,(cols,rows))
    
    image = pre_process_image(image)
    
    #image = cv2.cvtColor(image, cv2.COLOR_BGR2YUV)
    #image = image[:,:,0]
    #image = cv2.resize(image, (img_resize,img_resize),interpolation = cv2.INTER_CUBIC)
    
    return image

def get_index_dict(y_train):
    # Returns indices of each label
    # Assumes that the labels are 0 to N-1
    dict_indices = {}
    ind_all = np.arange(len(y_train))

    for i in range(len(np.unique(y_train))):
        ind_i = ind_all[y_train == i]
        dict_indices[i] = ind_i
        #print(ind_i)
    return dict_indices

def gen_transformed_data(X_train,y_train,N_classes,n_each,ang_range,shear_range,trans_range,randomize_Var):
   
   dict_indices = get_index_dict(y_train)
   n_class = len(np.unique(y_train))
   X_arr = []
   Y_arr = []
   for i in range(n_class):
       len_i = len(dict_indices[i])
       ind_rand = np.random.randint(0,len_i,n_each)
       ind_dict_class  = dict_indices[i]
       
       for i_n in range(n_each):
           img_trf = transform_image(X_train[ind_dict_class[ind_rand[i_n]]],
                                     ang_range,shear_range,trans_range)
           X_arr.append(img_trf)
           Y_arr.append(i)
           
   X_arr = np.array(X_arr,dtype = np.float32())
   Y_arr = np.array(Y_arr,dtype = np.float32())
   
   if (randomize_Var == 1):
       len_arr = np.arange(len(Y_arr))
       np.random.shuffle(len_arr)
       X_arr[len_arr] = X_arr
       Y_arr[len_arr] = Y_arr
       
   labels_arr = OHE_labels(Y_arr,43)

   return X_arr,Y_arr,labels_arr


def gen_extra_data(X_train,y_train,N_classes,n_each,ang_range,shear_range,trans_range,randomize_Var):
    
    dict_indices = get_index_dict(y_train)
    n_class = len(np.unique(y_train)) 
    X_arr = []
    Y_arr = []
    n_train = len(X_train)
    for i in range(n_train):
        #X_arr.append(transform_image(X_train[i],0,0,0)/255.)
        #Y_arr.append(y_train[i])
        for i_n in range(n_each):
            img_trf = transform_image(X_train[i],
                                      ang_range,shear_range,trans_range)
            X_arr.append(img_trf)
            Y_arr.append(y_train[i])
            
    X_arr = np.array(X_arr,dtype = np.float32())
    Y_arr = np.array(Y_arr,dtype = np.float32())
    
    if (randomize_Var == 1):
        len_arr = np.arange(len(Y_arr))
        np.random.shuffle(len_arr)
        X_arr[len_arr] = X_arr
        Y_arr[len_arr] = Y_arr
        
    labels_arr = OHE_labels(Y_arr,43)

    return X_arr,Y_arr,labels_arr