![\"Image](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/Image-classification-project.gif\")
<\/a><\/p>\nWhat is Image Classification<\/h2>\n\n- The classification problem is to categorize all the pixels of a digital image into one of the defined classes.<\/li>\n
- Image classification is the most critical use case in digital image analysis.<\/li>\n
- Image classification is an application of both supervised classification and unsupervised classification.\n
\n- In supervised classification, we select samples for each target class. We train our neural network<\/a> on these target class samples and then classify new samples.<\/li>\n
- In unsupervised classification, we group the sample images into clusters of images having similar properties. Then, we classify each cluster into our intended classes.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n
About Image Classification Dataset<\/h3>\n
CIFAR-10 is a very popular computer vision dataset. This dataset is well studied in many types of deep learning research for object recognition.<\/p>\n
This dataset consists of 60,000 images divided into 10 target classes, with each category containing 6000 images of shape 32*32.\u00a0This dataset contains images of low resolution (32*32), which allows researchers to try new algorithms. The 10 different classes of this dataset are:<\/p>\n
\n- Airplane<\/li>\n
- Car<\/li>\n
- Bird<\/li>\n
- Cat<\/li>\n
- Deer<\/li>\n
- Dog<\/li>\n
- Frog<\/li>\n
- Horse<\/li>\n
- Ship<\/li>\n
- Truck<\/li>\n<\/ol>\n
CIFAR-10 dataset is already available in the datasets module of Keras. We do not need to download it; we can directly import it from keras.datasets.<\/p>\n
Project Prerequisites:<\/h3>\n
The prerequisite to develop\u00a0and execute image classification project is\u00a0Keras and Tensorflow installation<\/a>.<\/p>\nSteps for image classification on CIFAR-10:<\/h3>\n
1. Load the dataset from keras datasets module<\/p>\n
from keras.datasets import cifar10\r\nimport matplotlib.pyplot as plt\r\n \r\n(train_X,train_Y),(test_X,test_Y)=cifar10.load_data()<\/pre>\n2.\u00a0Plot some images from the dataset to visualize the dataset<\/p>\n
n=6\r\nplt.figure(figsize=(20,10))\r\nfor i in range(n):\r\nplt.subplot(330+1+i)\r\nplt.imshow(train_X[i])\r\nplt.show()<\/pre>\n![\"image](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/visualize-dataset.png\")
<\/a><\/p>\n3. Import the required layers and modules to create our convolution neural net architecture<\/p>\n
from keras.models import Sequential\r\nfrom keras.layers import Dense\r\nfrom keras.layers import Dropout\r\nfrom keras.layers import Flatten\r\nfrom keras.constraints import maxnorm\r\nfrom keras.optimizers import SGD\r\nfrom keras.layers.convolutional import Conv2D\r\nfrom keras.layers.convolutional import MaxPooling2D\r\nfrom keras.utils import np_utils<\/pre>\n4. Convert the pixel values of the dataset to float type and then normalize the dataset<\/p>\n
train_x=train_X.astype('float32')\r\ntest_X=test_X.astype('float32')\r\n \r\ntrain_X=train_X\/255.0\r\ntest_X=test_X\/255.0<\/pre>\n5. Now perform the one-hot encoding for target classes<\/p>\n
train_Y=np_utils.to_categorical(train_Y)\r\ntest_Y=np_utils.to_categorical(test_Y)\r\n \r\nnum_classes=test_Y.shape[1]<\/pre>\n6.\u00a0Create the sequential model and add the layers<\/p>\n
model=Sequential()\r\nmodel.add(Conv2D(32,(3,3),input_shape=(32,32,3),\r\n padding='same',activation='relu',\r\n kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.2))\r\nmodel.add(Conv2D(32,(3,3),activation='relu',padding='same',kernel_constraint=maxnorm(3)))\r\nmodel.add(MaxPooling2D(pool_size=(2,2)))\r\nmodel.add(Flatten())\r\nmodel.add(Dense(512,activation='relu',kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.5))\r\nmodel.add(Dense(num_classes, activation='softmax'))<\/pre>\n![\"model](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/model-architecture.png\")
<\/a><\/p>\n7.\u00a0Configure the optimizer and compile the model<\/p>\n
sgd=SGD(lr=0.01,momentum=0.9,decay=(0.01\/25),nesterov=Fals)\r\n \r\nmodel.compile(loss='categorical_crossentropy',\r\n optimizer=sgd,\r\n metrics=['accuracy'])<\/pre>\n8.\u00a0View the model summary for better understanding of model architecture<\/p>\n
model.summary()<\/pre>\n![\"model](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/model-summary.png\")
<\/a><\/p>\n9.\u00a0Train the model<\/p>\n
model.fit(train_X,train_Y,\r\n validation_data=(test_X,test_Y),\r\n epochs=10,batch_size=32)<\/pre>\n![\"train](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/train-model.png\")
<\/a><\/p>\n10. Calculate its accuracy on testing data<\/p>\n
_,acc=model.evaluate(test_X,test_Y)\r\nprint(acc*100)<\/pre>\n11. Save the model<\/p>\n
model.save(\"model1_cifar_10epoch.h5\")<\/pre>\n12.\u00a0Make a dictionary to map to the output classes and make predictions from the model<\/p>\n
results={\r\n 0:'aeroplane',\r\n 1:'automobile',\r\n 2:'bird',\r\n 3:'cat',\r\n 4:'deer',\r\n 5:'dog',\r\n 6:'frog',\r\n 7:'horse',\r\n 8:'ship',\r\n 9:'truck'\r\n}\r\nfrom PIL import Image\r\nimport numpy as np\r\nim=Image.open(\"__image_path__\")\r\n# the input image is required to be in the shape of dataset, i.e (32,32,3)\r\n \r\nim=im.resize((32,32))\r\nim=np.expand_dims(im,axis=0)\r\nim=np.array(im)\r\npred=model.predict_classes([im])[0]\r\nprint(pred,results[pred])<\/pre>\n![\"model](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/05\/model-prediction.png\")
<\/a><\/p>\nYou can test the result on your custom image input. To improve accuracy, try increasing the epoch count to 25 for training.<\/p>\n
Image Classification Project GUI<\/h3>\n
- \n
- In supervised classification, we select samples for each target class. We train our neural network<\/a> on these target class samples and then classify new samples.<\/li>\n
- In unsupervised classification, we group the sample images into clusters of images having similar properties. Then, we classify each cluster into our intended classes.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n
About Image Classification Dataset<\/h3>\n
CIFAR-10 is a very popular computer vision dataset. This dataset is well studied in many types of deep learning research for object recognition.<\/p>\n
This dataset consists of 60,000 images divided into 10 target classes, with each category containing 6000 images of shape 32*32.\u00a0This dataset contains images of low resolution (32*32), which allows researchers to try new algorithms. The 10 different classes of this dataset are:<\/p>\n
- \n
- Airplane<\/li>\n
- Car<\/li>\n
- Bird<\/li>\n
- Cat<\/li>\n
- Deer<\/li>\n
- Dog<\/li>\n
- Frog<\/li>\n
- Horse<\/li>\n
- Ship<\/li>\n
- Truck<\/li>\n<\/ol>\n
CIFAR-10 dataset is already available in the datasets module of Keras. We do not need to download it; we can directly import it from keras.datasets.<\/p>\n
Project Prerequisites:<\/h3>\n
The prerequisite to develop\u00a0and execute image classification project is\u00a0Keras and Tensorflow installation<\/a>.<\/p>\n
Steps for image classification on CIFAR-10:<\/h3>\n
1. Load the dataset from keras datasets module<\/p>\n
from keras.datasets import cifar10\r\nimport matplotlib.pyplot as plt\r\n \r\n(train_X,train_Y),(test_X,test_Y)=cifar10.load_data()<\/pre>\n
2.\u00a0Plot some images from the dataset to visualize the dataset<\/p>\n
n=6\r\nplt.figure(figsize=(20,10))\r\nfor i in range(n):\r\nplt.subplot(330+1+i)\r\nplt.imshow(train_X[i])\r\nplt.show()<\/pre>\n
<\/a><\/p>\n3. Import the required layers and modules to create our convolution neural net architecture<\/p>\n
from keras.models import Sequential\r\nfrom keras.layers import Dense\r\nfrom keras.layers import Dropout\r\nfrom keras.layers import Flatten\r\nfrom keras.constraints import maxnorm\r\nfrom keras.optimizers import SGD\r\nfrom keras.layers.convolutional import Conv2D\r\nfrom keras.layers.convolutional import MaxPooling2D\r\nfrom keras.utils import np_utils<\/pre>\n
4. Convert the pixel values of the dataset to float type and then normalize the dataset<\/p>\n
train_x=train_X.astype('float32')\r\ntest_X=test_X.astype('float32')\r\n \r\ntrain_X=train_X\/255.0\r\ntest_X=test_X\/255.0<\/pre>\n5. Now perform the one-hot encoding for target classes<\/p>\n
train_Y=np_utils.to_categorical(train_Y)\r\ntest_Y=np_utils.to_categorical(test_Y)\r\n \r\nnum_classes=test_Y.shape[1]<\/pre>\n
6.\u00a0Create the sequential model and add the layers<\/p>\n
model=Sequential()\r\nmodel.add(Conv2D(32,(3,3),input_shape=(32,32,3),\r\n padding='same',activation='relu',\r\n kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.2))\r\nmodel.add(Conv2D(32,(3,3),activation='relu',padding='same',kernel_constraint=maxnorm(3)))\r\nmodel.add(MaxPooling2D(pool_size=(2,2)))\r\nmodel.add(Flatten())\r\nmodel.add(Dense(512,activation='relu',kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.5))\r\nmodel.add(Dense(num_classes, activation='softmax'))<\/pre>\n
<\/a><\/p>\n7.\u00a0Configure the optimizer and compile the model<\/p>\n
sgd=SGD(lr=0.01,momentum=0.9,decay=(0.01\/25),nesterov=Fals)\r\n \r\nmodel.compile(loss='categorical_crossentropy',\r\n optimizer=sgd,\r\n metrics=['accuracy'])<\/pre>\n
8.\u00a0View the model summary for better understanding of model architecture<\/p>\n
model.summary()<\/pre>\n
<\/a><\/p>\n9.\u00a0Train the model<\/p>\n
model.fit(train_X,train_Y,\r\n validation_data=(test_X,test_Y),\r\n epochs=10,batch_size=32)<\/pre>\n
<\/a><\/p>\n10. Calculate its accuracy on testing data<\/p>\n
_,acc=model.evaluate(test_X,test_Y)\r\nprint(acc*100)<\/pre>\n
11. Save the model<\/p>\n
model.save(\"model1_cifar_10epoch.h5\")<\/pre>\n
12.\u00a0Make a dictionary to map to the output classes and make predictions from the model<\/p>\n
results={\r\n 0:'aeroplane',\r\n 1:'automobile',\r\n 2:'bird',\r\n 3:'cat',\r\n 4:'deer',\r\n 5:'dog',\r\n 6:'frog',\r\n 7:'horse',\r\n 8:'ship',\r\n 9:'truck'\r\n}\r\nfrom PIL import Image\r\nimport numpy as np\r\nim=Image.open(\"__image_path__\")\r\n# the input image is required to be in the shape of dataset, i.e (32,32,3)\r\n \r\nim=im.resize((32,32))\r\nim=np.expand_dims(im,axis=0)\r\nim=np.array(im)\r\npred=model.predict_classes([im])[0]\r\nprint(pred,results[pred])<\/pre>\n<\/a><\/p>\nYou can test the result on your custom image input. To improve accuracy, try increasing the epoch count to 25 for training.<\/p>\n
Image Classification Project GUI<\/h3>\n
- In unsupervised classification, we group the sample images into clusters of images having similar properties. Then, we classify each cluster into our intended classes.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n