【深度学习】图片分类CNN模板

构建一个字母ABC的手写识别网络,

要求给出算法误差收敛曲线，所给程序要有图片导入接口。

其中A，B，C都代表label，三个文件夹存在具体的图片。只要是这样类型的，直接套下面模板。

import os
import cv2
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import tensorflow as tf
import tqdmfrom tensorflow import keras
from keras import Input, Model, Sequential
from tensorflow.keras.regularizers import l2
from keras.layers import Dense, Flatten, InputLayer, Reshape, BatchNormalization, Dropout, Conv2D, MaxPooling2D
from tensorflow.keras.utils import plot_model%matplotlib inline

data_dir = './data'

# Dict of labels
categories = {
'A': 0,
'B': 1,
'C': 2
}

def load_images(images_folder, img_size = (128,128), scale=False):# Store paths to imagesimage_path = []for dirname, _, filenames in os.walk(images_folder):for filename in filenames:image_path.append(os.path.join(dirname, filename))print("There are {} images in {}".format(len(image_path), images_folder))# Load images and associated labelsimages = []labels = []for path in tqdm.tqdm(image_path):img = cv2.imread(path)    img = cv2.resize(img, img_size) # Resize the imagesimg = np.array(img)images.append(img)labels.append(categories[path.split('/')[-2]]) # last folder before the image name is the categoryimages = np.array(images)  images = images.astype(np.int64)if scale:images = images/255 # scalereturn image_path, images, np.asarray(labels)

img_size = (128,128)
image_path, images, labels = load_images(data_dir, img_size=img_size)# Resize
# images = np.array(images).reshape(-1,128,128,1)
images.shape

There are 600 images in ./data100%|██████████| 600/600 [00:03<00:00, 183.15it/s](600, 128, 128, 3)

# 查看图片
plt.figure(figsize=(10,10))
random_inds = np.random.choice(len(image_path),36)
for i in range(36):plt.subplot(6,6,i+1)plt.xticks([])plt.yticks([])plt.grid(False)image_ind = random_inds[i]plt.imshow(np.squeeze(images[image_ind]), cmap=plt.cm.binary)label = list(categories.keys())[list(categories.values()).index(labels[image_ind])]plt.title(label)

labels_df = pd.DataFrame(labels)
labels_df.value_counts()

2    201
0    201
1    198
dtype: int64

dataset=[]
dataname=[]
count=0
for name in tqdm(os.listdir(data_dir)):path=os.path.join(data_dir,name)for im in os.listdir(path):image=cv2.imread(os.path.join(path,im))image2=np.resize(image,(50,50,3))dataset+=[image2]dataname+=[count]count=count+1

100%|██████████| 3/3 [00:03<00:00,  1.06s/it]

data=np.array(dataset)
dataname=np.array(dataname)

data[0].shape

(50, 50, 3)

print(pd.Series(dataname).value_counts())

1    202
2    201
0    198
dtype: int64

len(categories)

from tensorflow.keras.layers import Dense, Conv2D, Flatten, Dropout, MaxPooling2D, BatchNormalizationdef build_cnn_model():cnn_model=tf.keras.Sequential([Conv2D(filters=32,kernel_size=(3,3),activation='relu',input_shape=images.shape[1:]),MaxPooling2D(2,2),BatchNormalization(),Dropout(0.4),Conv2D(filters=64,kernel_size=(3,3),activation='relu', padding='same'),Conv2D(filters=64,kernel_size=(3,3),activation='relu', padding='same'),MaxPooling2D((2,2)),BatchNormalization(),Dropout(0.4),Conv2D(filters=128,kernel_size=(3,3),activation='relu', padding='same'),Conv2D(filters=128,kernel_size=(3,3),activation='relu', padding='same'),MaxPooling2D(2,2),BatchNormalization(),Dropout(0.4),Conv2D(filters=256,kernel_size=(3,3),activation='relu', padding='same'),Conv2D(filters=256,kernel_size=(3,3),activation='relu', padding='same'),MaxPooling2D(2,2),BatchNormalization(),Dropout(0.4),Conv2D(filters=128,kernel_size=(3,3),activation='relu', padding='same'),Conv2D(filters=128,kernel_size=(3,3),activation='relu', padding='same'),MaxPooling2D(2,2),BatchNormalization(),Dropout(0.4),Conv2D(filters=64,kernel_size=(3,3),activation='relu', padding='same'),Conv2D(filters=64,kernel_size=(3,3),activation='relu', padding='same'),MaxPooling2D((2,2)),BatchNormalization(),Dropout(0.4),Flatten(),Dense(units=len(categories),activation='softmax')])return cnn_modelmodel = build_cnn_model()
# Initialize the model by passing some data through
model.predict(images[[0]])
# Print the summary of the layers in the model.
print(model.summary())

Model: "sequential_3"
_________________________________________________________________Layer (type)                Output Shape              Param #
=================================================================conv2d_6 (Conv2D)           (None, 126, 126, 32)      896       max_pooling2d_6 (MaxPooling  (None, 63, 63, 32)       0         2D)                                                             batch_normalization (BatchN  (None, 63, 63, 32)       128       ormalization)                                                   dropout (Dropout)           (None, 63, 63, 32)        0         conv2d_7 (Conv2D)           (None, 63, 63, 64)        18496     conv2d_8 (Conv2D)           (None, 63, 63, 64)        36928     max_pooling2d_7 (MaxPooling  (None, 31, 31, 64)       0         2D)                                                             batch_normalization_1 (Batc  (None, 31, 31, 64)       256       hNormalization)                                                 dropout_1 (Dropout)         (None, 31, 31, 64)        0         conv2d_9 (Conv2D)           (None, 31, 31, 128)       73856     conv2d_10 (Conv2D)          (None, 31, 31, 128)       147584    max_pooling2d_8 (MaxPooling  (None, 15, 15, 128)      0         2D)                                                             batch_normalization_2 (Batc  (None, 15, 15, 128)      512       hNormalization)                                                 dropout_2 (Dropout)         (None, 15, 15, 128)       0         conv2d_11 (Conv2D)          (None, 15, 15, 256)       295168    conv2d_12 (Conv2D)          (None, 15, 15, 256)       590080    max_pooling2d_9 (MaxPooling  (None, 7, 7, 256)        0         2D)                                                             batch_normalization_3 (Batc  (None, 7, 7, 256)        1024      hNormalization)                                                 dropout_3 (Dropout)         (None, 7, 7, 256)         0         conv2d_13 (Conv2D)          (None, 7, 7, 128)         295040    conv2d_14 (Conv2D)          (None, 7, 7, 128)         147584    max_pooling2d_10 (MaxPoolin  (None, 3, 3, 128)        0         g2D)                                                            batch_normalization_4 (Batc  (None, 3, 3, 128)        512       hNormalization)                                                 dropout_4 (Dropout)         (None, 3, 3, 128)         0         conv2d_15 (Conv2D)          (None, 3, 3, 64)          73792     conv2d_16 (Conv2D)          (None, 3, 3, 64)          36928     max_pooling2d_11 (MaxPoolin  (None, 1, 1, 64)         0         g2D)                                                            batch_normalization_5 (Batc  (None, 1, 1, 64)         256       hNormalization)                                                 dropout_5 (Dropout)         (None, 1, 1, 64)          0         flatten_1 (Flatten)         (None, 64)                0         dense_6 (Dense)             (None, 3)                 195       =================================================================
Total params: 1,719,235
Trainable params: 1,717,891
Non-trainable params: 1,344
_________________________________________________________________
None

tf.keras.utils.plot_model(model, show_shapes=True)

from tensorflow.keras.utils import to_categorical
from sklearn.preprocessing import LabelEncoder
from sklearn.utils import shuffle
le = LabelEncoder()
labels = le.fit_transform(labels)
labels = to_categorical(labels)
labels[:10]

array([[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.],[0., 1., 0.]], dtype=float32)

model.compile(optimizer = "adam", loss = "binary_crossentropy", metrics = ["accuracy"])

checkpoint_filepath = '/checkpoint.hdf5'
model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_filepath,save_weights_only=True,monitor='val_accuracy',mode='max',save_best_only=True,save_freq=500)

from tensorflow.keras.callbacks import EarlyStopping
early_stopping = EarlyStopping(monitor='val_accuracy', patience=10, min_delta=0.001, mode='max',restore_best_weights=True
)

datagen = ImageDataGenerator(horizontal_flip=True,vertical_flip=True,rotation_range=20,zoom_range=0.2,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.1,fill_mode="nearest")

from tensorflow.keras.callbacks import ReduceLROnPlateau
reducelr = ReduceLROnPlateau(monitor = "val_accuracy",factor = 0.3, patience = 3,min_delta = 0.001,mode = 'auto',verbose=1)

from sklearn.model_selection import train_test_split
# Train, validation and test splitX_train, X_test, y_train, y_test = train_test_split(images, labels, test_size=0.10, random_state=7)X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.20, random_state=1)

print("*-*-*-*-*-*")
print("Train")
print(X_train.shape)
print(y_train.shape)print("*-*-*-*-*-*")
print("Validation")
print(X_val.shape)
print(y_val.shape)print("*-*-*-*-*-*")
print("Test")
print(X_test.shape)
print(y_test.shape)

*-*-*-*-*-*
Train
(432, 128, 128, 3)
(432, 3)
*-*-*-*-*-*
Validation
(108, 128, 128, 3)
(108, 3)
*-*-*-*-*-*
Test
(60, 128, 128, 3)
(60, 3)

history = model.fit(X_train, y_train, batch_size = 32, epochs = 100, verbose = 1, validation_data = (X_val, y_val),callbacks=[model_checkpoint_callback, early_stopping, reducelr])

Epoch 1/1000
14/14 [==============================] - 8s 238ms/step - loss: 0.8036 - accuracy: 0.3588 - val_loss: 6.2921 - val_accuracy: 0.2963 - lr: 0.0010
Epoch 2/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.8162 - accuracy: 0.3796 - val_loss: 5.2361 - val_accuracy: 0.2963 - lr: 0.0010
Epoch 3/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.7190 - accuracy: 0.4537 - val_loss: 1.3893 - val_accuracy: 0.3333 - lr: 0.0010
Epoch 4/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.6875 - accuracy: 0.4792 - val_loss: 0.7386 - val_accuracy: 0.3519 - lr: 0.0010
Epoch 5/1000
14/14 [==============================] - 1s 100ms/step - loss: 0.6144 - accuracy: 0.5949 - val_loss: 0.7014 - val_accuracy: 0.4259 - lr: 0.0010
Epoch 6/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.5156 - accuracy: 0.7060 - val_loss: 0.7592 - val_accuracy: 0.4537 - lr: 0.0010
Epoch 7/1000
14/14 [==============================] - 1s 96ms/step - loss: 0.4904 - accuracy: 0.7384 - val_loss: 0.7034 - val_accuracy: 0.5370 - lr: 0.0010
Epoch 8/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.3854 - accuracy: 0.7940 - val_loss: 0.6092 - val_accuracy: 0.5556 - lr: 0.0010
Epoch 9/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.3313 - accuracy: 0.8241 - val_loss: 0.5192 - val_accuracy: 0.6389 - lr: 0.0010
Epoch 10/1000
14/14 [==============================] - 1s 93ms/step - loss: 0.2873 - accuracy: 0.8519 - val_loss: 0.5089 - val_accuracy: 0.6111 - lr: 0.0010
Epoch 11/1000
14/14 [==============================] - 1s 96ms/step - loss: 0.2346 - accuracy: 0.8981 - val_loss: 0.4359 - val_accuracy: 0.6852 - lr: 0.0010
Epoch 12/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.2238 - accuracy: 0.8819 - val_loss: 0.4404 - val_accuracy: 0.6481 - lr: 0.0010
Epoch 13/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.1954 - accuracy: 0.8912 - val_loss: 0.4215 - val_accuracy: 0.7500 - lr: 0.0010
Epoch 14/1000
14/14 [==============================] - 1s 100ms/step - loss: 0.1792 - accuracy: 0.9051 - val_loss: 0.1971 - val_accuracy: 0.9074 - lr: 0.0010
Epoch 15/1000
14/14 [==============================] - 1s 96ms/step - loss: 0.1608 - accuracy: 0.9144 - val_loss: 0.2836 - val_accuracy: 0.8056 - lr: 0.0010
Epoch 16/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.1447 - accuracy: 0.9398 - val_loss: 0.2867 - val_accuracy: 0.7500 - lr: 0.0010
Epoch 17/1000
14/14 [==============================] - ETA: 0s - loss: 0.1215 - accuracy: 0.9375
Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.0003000000142492354.
14/14 [==============================] - 1s 95ms/step - loss: 0.1215 - accuracy: 0.9375 - val_loss: 0.1474 - val_accuracy: 0.9074 - lr: 0.0010
Epoch 18/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.1023 - accuracy: 0.9537 - val_loss: 0.1186 - val_accuracy: 0.9352 - lr: 3.0000e-04
Epoch 19/1000
14/14 [==============================] - 1s 101ms/step - loss: 0.0992 - accuracy: 0.9606 - val_loss: 0.1074 - val_accuracy: 0.9444 - lr: 3.0000e-04
Epoch 20/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0837 - accuracy: 0.9676 - val_loss: 0.0917 - val_accuracy: 0.9444 - lr: 3.0000e-04
Epoch 21/1000
14/14 [==============================] - 1s 98ms/step - loss: 0.0788 - accuracy: 0.9699 - val_loss: 0.0877 - val_accuracy: 0.9444 - lr: 3.0000e-04
Epoch 22/1000
14/14 [==============================] - ETA: 0s - loss: 0.0809 - accuracy: 0.9722
Epoch 00022: ReduceLROnPlateau reducing learning rate to 9.000000427477062e-05.
14/14 [==============================] - 1s 95ms/step - loss: 0.0809 - accuracy: 0.9722 - val_loss: 0.0897 - val_accuracy: 0.9444 - lr: 3.0000e-04
Epoch 23/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.0677 - accuracy: 0.9792 - val_loss: 0.0834 - val_accuracy: 0.9537 - lr: 9.0000e-05
Epoch 24/1000
14/14 [==============================] - 1s 93ms/step - loss: 0.0741 - accuracy: 0.9722 - val_loss: 0.0771 - val_accuracy: 0.9537 - lr: 9.0000e-05
Epoch 25/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0672 - accuracy: 0.9815 - val_loss: 0.0733 - val_accuracy: 0.9537 - lr: 9.0000e-05
Epoch 26/1000
14/14 [==============================] - ETA: 0s - loss: 0.0595 - accuracy: 0.9838
Epoch 00026: ReduceLROnPlateau reducing learning rate to 2.700000040931627e-05.
14/14 [==============================] - 1s 95ms/step - loss: 0.0595 - accuracy: 0.9838 - val_loss: 0.0694 - val_accuracy: 0.9537 - lr: 9.0000e-05
Epoch 27/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0631 - accuracy: 0.9838 - val_loss: 0.0699 - val_accuracy: 0.9537 - lr: 2.7000e-05
Epoch 28/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.0591 - accuracy: 0.9861 - val_loss: 0.0705 - val_accuracy: 0.9537 - lr: 2.7000e-05
Epoch 29/1000
14/14 [==============================] - ETA: 0s - loss: 0.0635 - accuracy: 0.9838
Epoch 00029: ReduceLROnPlateau reducing learning rate to 8.100000013655517e-06.
14/14 [==============================] - 1s 95ms/step - loss: 0.0635 - accuracy: 0.9838 - val_loss: 0.0697 - val_accuracy: 0.9444 - lr: 2.7000e-05
Epoch 30/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.0643 - accuracy: 0.9792 - val_loss: 0.0687 - val_accuracy: 0.9444 - lr: 8.1000e-06
Epoch 31/1000
14/14 [==============================] - 1s 100ms/step - loss: 0.0768 - accuracy: 0.9745 - val_loss: 0.0665 - val_accuracy: 0.9537 - lr: 8.1000e-06
Epoch 32/1000
14/14 [==============================] - ETA: 0s - loss: 0.0645 - accuracy: 0.9861
Epoch 00032: ReduceLROnPlateau reducing learning rate to 2.429999949526973e-06.
14/14 [==============================] - 1s 95ms/step - loss: 0.0645 - accuracy: 0.9861 - val_loss: 0.0656 - val_accuracy: 0.9537 - lr: 8.1000e-06
Epoch 33/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.0635 - accuracy: 0.9792 - val_loss: 0.0645 - val_accuracy: 0.9630 - lr: 2.4300e-06
Epoch 34/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.0606 - accuracy: 0.9838 - val_loss: 0.0636 - val_accuracy: 0.9630 - lr: 2.4300e-06
Epoch 35/1000
14/14 [==============================] - 1s 95ms/step - loss: 0.0620 - accuracy: 0.9907 - val_loss: 0.0628 - val_accuracy: 0.9630 - lr: 2.4300e-06
Epoch 36/10009/14 [==================>...........] - ETA: 0s - loss: 0.0729 - accuracy: 0.9826WARNING:tensorflow:Can save best model only with val_accuracy available, skipping.
14/14 [==============================] - ETA: 0s - loss: 0.0682 - accuracy: 0.9861
Epoch 00036: ReduceLROnPlateau reducing learning rate to 7.289999985005124e-07.
14/14 [==============================] - 1s 95ms/step - loss: 0.0682 - accuracy: 0.9861 - val_loss: 0.0622 - val_accuracy: 0.9630 - lr: 2.4300e-06
Epoch 37/1000
14/14 [==============================] - 1s 96ms/step - loss: 0.0573 - accuracy: 0.9907 - val_loss: 0.0613 - val_accuracy: 0.9630 - lr: 7.2900e-07
Epoch 38/1000
14/14 [==============================] - 1s 97ms/step - loss: 0.0575 - accuracy: 0.9931 - val_loss: 0.0607 - val_accuracy: 0.9722 - lr: 7.2900e-07
Epoch 39/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0622 - accuracy: 0.9769 - val_loss: 0.0600 - val_accuracy: 0.9722 - lr: 7.2900e-07
Epoch 40/1000
14/14 [==============================] - 1s 96ms/step - loss: 0.0660 - accuracy: 0.9838 - val_loss: 0.0594 - val_accuracy: 0.9722 - lr: 7.2900e-07
Epoch 41/1000
14/14 [==============================] - ETA: 0s - loss: 0.0614 - accuracy: 0.9884
Epoch 00041: ReduceLROnPlateau reducing learning rate to 2.1870000637136398e-07.
14/14 [==============================] - 1s 95ms/step - loss: 0.0614 - accuracy: 0.9884 - val_loss: 0.0591 - val_accuracy: 0.9722 - lr: 7.2900e-07
Epoch 42/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0605 - accuracy: 0.9792 - val_loss: 0.0583 - val_accuracy: 0.9722 - lr: 2.1870e-07
Epoch 43/1000
14/14 [==============================] - 1s 99ms/step - loss: 0.0529 - accuracy: 0.9954 - val_loss: 0.0582 - val_accuracy: 0.9722 - lr: 2.1870e-07
Epoch 44/1000
14/14 [==============================] - ETA: 0s - loss: 0.0500 - accuracy: 0.9884
Epoch 00044: ReduceLROnPlateau reducing learning rate to 6.561000276406048e-08.
14/14 [==============================] - 1s 95ms/step - loss: 0.0500 - accuracy: 0.9884 - val_loss: 0.0580 - val_accuracy: 0.9722 - lr: 2.1870e-07
Epoch 45/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0613 - accuracy: 0.9861 - val_loss: 0.0581 - val_accuracy: 0.9722 - lr: 6.5610e-08
Epoch 46/1000
14/14 [==============================] - 1s 94ms/step - loss: 0.0672 - accuracy: 0.9861 - val_loss: 0.0572 - val_accuracy: 0.9722 - lr: 6.5610e-08
Epoch 47/1000
14/14 [==============================] - ETA: 0s - loss: 0.0511 - accuracy: 0.9931
Epoch 00047: ReduceLROnPlateau reducing learning rate to 1.9683000829218145e-08.
14/14 [==============================] - 1s 96ms/step - loss: 0.0511 - accuracy: 0.9931 - val_loss: 0.0574 - val_accuracy: 0.9722 - lr: 6.5610e-08
Epoch 48/1000
14/14 [==============================] - 1s 99ms/step - loss: 0.0622 - accuracy: 0.9861 - val_loss: 0.0570 - val_accuracy: 0.9722 - lr: 1.9683e-08

plt.plot(history.history["accuracy"])
plt.plot(history.history["val_accuracy"])
plt.title("Model accuracy")
plt.ylabel("Accuracy")
plt.xlabel("Epoch")
plt.legend(["Train", "Test"], loc = "upper left")
plt.show()

plt.plot(history.history["loss"])
plt.plot(history.history["val_loss"])
plt.title("Model loss")
plt.ylabel("Loss")
plt.xlabel("Epoch")
plt.legend(["Train", "Test"], loc = "upper left")
plt.show()

def predict_class(img):# Resizeimg = img.reshape(1,128,128,3)# Predictpredictions = model.predict(img)true_prediction = [tf.argmax(pred) for pred in predictions]true_prediction = np.array(true_prediction)# Return label corresponding to predicted indexreturn list(categories.keys())[list(categories.values()).index(true_prediction)]

# Predict on test set
y_pred = model.predict(X_test)

# From categorical outputs to discrete values
y_pred_ = [np.argmax(y) for y in y_pred]
y_test_ = [np.argmax(y) for y in y_test]

from sklearn.metrics import classification_report
print(classification_report(y_test_, y_pred_))

precision    recall  f1-score   support0       1.00      0.96      0.98        251       0.85      1.00      0.92        112       1.00      0.96      0.98        24accuracy                           0.97        60macro avg       0.95      0.97      0.96        60
weighted avg       0.97      0.97      0.97        60

plt.figure(figsize=(10,10))
random_inds = np.random.choice(X_test.shape[0],36)
for i in range(36):plt.subplot(6,6,i+1)plt.xticks([])plt.yticks([])plt.grid(False)image_ind = random_inds[i]plt.imshow(np.squeeze(X_test[image_ind]), cmap=plt.cm.binary)# Predict and get labellabel = predict_class(X_test[image_ind])plt.xlabel(label)

model.save("model.h5")

# 识别接口
def predict(path,model_str,img_size = (128,128)):new_model = tf.keras.models.load_model(model_str)img = cv2.imread(path)img = cv2.resize(img, img_size) # Resize the imagesimg = np.array(img)    # Resizeimg = img.reshape(1,128,128,3)# Predictpredictions = new_model.predict(img)true_prediction = [tf.argmax(pred) for pred in predictions]true_prediction = np.array(true_prediction)# Return label corresponding to predicted indexreturn list(categories.keys())[list(categories.values()).index(true_prediction)]predict("./data/A/051.jpg","model.h5")

'A'

predict("./data/B/048.jpg","model.h5")

'B'

predict("./data/C/050.jpg","model.h5")

'C'

往期精彩回顾适合初学者入门人工智能的路线及资料下载中国大学慕课《机器学习》（黄海广主讲）机器学习及深度学习笔记等资料打印机器学习在线手册深度学习笔记专辑《统计学习方法》的代码复现专辑
AI基础下载本站qq群955171419，加入微信群请扫码：

【深度学习】图片分类CNN模板相关推荐

深度学习图片分类CNN模板
构建一个字母ABC的手写识别网络, 要求给出算法误差收敛曲线,所给程序要有图片导入接口. 其中A,B,C都代表label,三个文件夹存在具体的图片.只要是这样类型的,直接套下面模板. import o ...
深度学习图片分类实战学习
开始记录学习深度学习的点点滴滴深度学习图片分类实战学习前言一.深度学习二.使用步骤 1. 自建网络模型 2. 进行深度学习的学习迁移注意事项前言随着人工智能的不断发展,这门技术也越来越重 ...
基于逻辑回归，支持向量机，朴素贝叶斯以及简单深度学习文本分类方法（BiLSTM、CNN）实现的中文情感分析，含数据集可直接运行
基于逻辑回归,支持向量机,朴素贝叶斯以及简单深度学习文本分类方法(BiLSTM.CNN)实现的中文情感分析,含数据集可直接运行完整代码下载地址:中文情感分析中文情感分析本项目旨在通过一个中文情感 ...
深度学习文本分类文献综述（翻译自Deep Learning Based Text Classification: A Comprehensive Review）
深度学习文本分类文献综述摘要介绍 1. 文本分类任务 2．文本分类中的深度模型 2.1 Feed-Forward Neural Networks 2.2 RNN-Based Models 2.3 ...
【NLP】深度学习文本分类｜模型代码技巧
文本分类是NLP的必备入门任务,在搜索.推荐.对话等场景中随处可见,并有情感分析.新闻分类.标签分类等成熟的研究分支和数据集. 本文主要介绍深度学习文本分类的常用模型原理.优缺点以及技巧,是「NLP入 ...
传统文本分类和基于深度学习文本分类
用深度学习(CNN RNN Attention)解决大规模文本分类问题 - 综述和实践近来在同时做一个应用深度学习解决淘宝商品的类目预测问题的项目,恰好硕士毕业时论文题目便是文本分类问题,趁此机会总 ...
(zhuan) 126 篇殿堂级深度学习论文分类整理从入门到应用
126 篇殿堂级深度学习论文分类整理从入门到应用 | 干货雷锋网作者: 三川 2017-03-02 18:40:00 查看源网址阅读数:66 如果你有非常大的决心从事深度学习,又不想在这一行 ...
Keras 搭建图片分类 CNN （卷积神经网络）
1. 导入keras from keras.models import Sequential from keras.layers import Conv2D, MaxPooling2D, Flatte ...
深度学习 | MATLAB实现CNN卷积神经网络结构及参数概述
深度学习 | MATLAB实现CNN卷积神经网络结构及参数概述目录深度学习 | MATLAB实现CNN卷积神经网络结构及参数概述基本介绍模型描述训练过程计算函数参考资料基本介绍卷积神 ...

【深度学习】图片分类CNN模板

【深度学习】图片分类CNN模板相关推荐

最新文章

热门文章