[AI] cifar10

 

cifar10 : mnist와 같이 머신러닝 연구에 사용되는 dataset 

 

 

day17)

1. mycifar10_1

import tensorflow as tf import numpy as np # 1. Fashion MNIST 데이터셋 임포트 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data() # 2. 데이터 전처리 train_images, test_images = train_images / 255.0, test_images / 255.0 # 3. 모델 구성 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(28, 28)), tf.keras.layers.Dense(512, activation='relu'), tf.keras.layers.Dense(10, activation='softmax') #10 : class 리스트의 종류 ]) # 4. 모델 컴파일 model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) # 5. 모델 훈련 model.fit(train_images, train_labels, epochs=5) # 6. 정확도 평가하기 loss, accuracy = model.evaluate(test_images, test_labels) print(loss, accuracy) # 7. 예측하기 predictions = model.predict(test_images) print(predictions[0]) print(np.argmax(predictions[0]))

2-1. mycifar10_2_hw

import tensorflow as tf import numpy as np import cv2 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data() # print("train_images",train_images.shape) # print("train_labels",train_labels.shape) # print("test_images",test_images.shape) # print("test_labels",test_labels.shape) # # print("train_images",train_images[0]) # print("train_labels",train_labels[0]) # print("test_images",test_images[0]) # print("test_labels",test_labels[0]) train_images, test_images = train_images/255.0, test_images/255.0 # **2: 이미지가 안 뜨는 현상 고치기 : or x_test256으로 바꾸기 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(28, 28)), tf.keras.layers.Dense(512, activation=tf.nn.relu), tf.keras.layers.Dense(10, activation=tf.nn.softmax) ]) model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) #accuracy가 중요함... 이게 높아질수록 안정된 데이터를 얻을 수 있다 model.fit(train_images, train_labels, epochs=5) print("test_images",test_images.shape) #뇌의 모양 확인 #예측하기 ★★★ 훈련 뒤 얼마나 예측을 잘 하는지, 훈련한것을 바탕으로 예측함 predict = model.predict(test_images) cnt_o = 0 cnt_x = 0 for i in range(10000): pred = np.argmax(predict[i]) #argmax: numpy 안에서 최대값의 index를 보여줌 (배열에는 못 쓰고 numpy에서만 사용) goog = train_labels[i] if pred == goog: cnt_o += 1 else: cnt_x += 1 print("i",i,"pred",pred,"goog",goog) # i번호_pred_goog.jpg로 fault 폴더에 저장하기(틀린 것들만 뽑아내기) print("{}_{}_{}".format(i,pred,goog)) cv2.imwrite('fault/{}_{}_{}.jpg'.format(i,pred,goog), test_images[i]*255.0) #255.0을 곱해야 이미지가 제대로 나옴 안 곱해주면 0만 들어가있는 상황임 **1 **4 :x_test256 # print("pred",np.argmax(predict[0])) #예측한 숫자, 배열 중에 가장 확률이 높은 index를 뽑아라 # print("goog",y_test[0]) #google이 준 숫자 print("cnt_o",cnt_o) print("cnt_x",cnt_x)

cnt_o = 0  cnt_x = 0 for i in range(10000):     pred = np.argmax(predict[i])  #argmax: numpy 안에서 최대값의 index를 보여줌 (배열에는 못 쓰고 numpy에서만 사용)      goog = train_labels[i]     if pred == goog:                 cnt_o += 1     else:         cnt_x += 1         print("i",i,"pred",pred,"goog",goog)         # 틀린 것들만 뽑아내기 : i번호_pred_goog.jpg로 fault 폴더에 저장하기         print("{}_{}_{}".format(i,pred,goog))         cv2.imwrite('fault/{}_{}_{}.jpg'.format(i,pred,goog), test_images[i]*255.0) #255.0을 곱해야 이미지가 제대로 나옴 안 곱해주면 0만 들어가있는 상황임

2-2. mycifar10_2 (강사님 코드)

import tensorflow as tf import numpy as np import cv2 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data() for idx,i in enumerate(train_images): label = train_labels[idx] cv2.imwrite('train/{}/{}.jpg'.format(label,idx), train_images[idx]) for idx,i in enumerate(test_images): label = test_labels[idx] cv2.imwrite('test/{}/{}.jpg'.format(label,idx), test_images[idx])

 

3.mycifar10_3_predict

import tensorflow as tf import numpy as np import cv2 # 0 : T-shirt/top # 1 : Trouser # 2 : Pullover # 3 : Dress # 4 : Coat # 5 : Sandal # 6 : Shirt # 7 : Sneaker # 8 : Bag # 9 : Ankel boot # 1. Fashion MNIST 데이터셋 임포트 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images_256, test_labels) = fashion_mnist.load_data() train_images, test_images = train_images/255.0, test_images_256/255.0 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(28, 28)), tf.keras.layers.Dense(512, activation=tf.nn.relu), tf.keras.layers.Dense(10, activation=tf.nn.softmax) ]) model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) model.fit(train_images, train_labels, epochs=5) predictions = model.predict(test_images) for idx, i in enumerate(predictions): idx_goog = test_labels[idx] idx_pred = np.argmax(predictions[idx]) print(idx,idx_pred,idx_goog) if idx_goog != idx_pred: cv2.imwrite('fault/{}_{}_{}.jpg'.format(i,idx_pred,idx_goog),test_images_256[idx])

predictions = model.predict(test_images) for idx, i in enumerate(predictions):     idx_goog = test_labels[idx]     idx_pred = np.argmax(predictions[idx])     print(idx,idx_pred,idx_goog)          if idx_goog != idx_pred:         cv2.imwrite('fault/{}_{}_{}.jpg'.format(i,idx_pred,idx_goog),test_images_256[idx])  : 틀린 이미지는 fault 폴더에 넣기

 

4. mycifar10_4_save

import tensorflow as tf import numpy as np import cv2 import tensorflow as tf import numpy as np # 1. Fashion MNIST 데이터셋 임포트 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images_256, test_labels) = fashion_mnist.load_data() # 2. 데이터 전처리 train_images, test_images = train_images / 255.0, test_images_256/255.0 # 3. 모델 구성 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(28, 28)), #한줄로만들기 tf.keras.layers.Dense(512, activation='relu'), tf.keras.layers.Dense(10, activation='softmax') #10가지 클래스 ]) # 4. 모델 컴파일 model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) # 5. 모델 훈련 model.fit(train_images, train_labels, epochs=10) model.save('fashion.h5')

model.save('fashion.h5')

 

5. mycifar10_5_load

import tensorflow as tf import numpy as np import cv2 import tensorflow as tf import numpy as np from keras.models import load_model import keras # 1. Fashion MNIST 데이터셋 임포트 fashion_mnist = tf.keras.datasets.fashion_mnist (train_images, train_labels), (test_images_256, test_labels) = fashion_mnist.load_data() # 2. 데이터 전처리 train_images, test_images = train_images/255.0, test_images_256/255.0 model = keras.models.load_model("fashion.h5") # 6. 정확도 평가하기 loss, accuracy = model.evaluate(test_images, test_labels) print(loss, accuracy) # # # 7. 예측하기 predictions = model.predict(test_images) print("idx_goog :",test_labels[0]) print("predictions[0]: ",predictions[0]) print("idx_pred:",np.argmax(predictions[0]))

model = keras.models.load_model("fashion.h5")

 

 

 

6. mycifar10_6_Internet

import tensorflow as tf import numpy as np import cv2 import tensorflow as tf import numpy as np from keras.models import load_model import keras classes = [ "T-shirt/top","Trouser","Pullover","Dress","Coat", "Sandal","Shirt","Sneaker","Bag","Ankel boot" ] img = cv2.imread('37.jpg') #numpy img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) resize_img = cv2.resize(img_gray, (28, 28))/255 # resize_img = 1 - resize_img #이미지 반전 resize_img = np.reshape(resize_img,(1,28,28)) model = keras.models.load_model("fashion.h5") # # 7. 예측하기 predictions = model.predict(resize_img) idx_pred = np.argmax(predictions[0]) print(idx_pred, classes[idx_pred])