TensorFlow 2.0 Tutorial for Beginners 13 - Multi-Label Image Classification on Movies Poster in CNN 

HERO

TY

Thank you so much for watching it. Please keep watching and happy Learning. Let us know if you need a lesson on any other topic.

@@KGPTalkie It's nice to watch this video. I wanted to know that how I can predict the rating of a image by poster ? if you can extend the same video with rating prediction will be very helpful.

Thank you so much for watching.

Thank you ❤️ 😍

I think using softmax as the output activation function solves most of multiclass porblem.

❤️ 😍

Thank you ❤️

sir, if possible please share the csv file in your GitHub or any other platform, it would be a great help.

Thanks for watching. I think you have to do it manually. Mostly data are prepared manually by someone.

@@KGPTalkie thanks

Hello, what can we do instead ?

@@olcaybolat3452 I know you commented this a year ago, but just incase you're still wondering (and for future viewers who are getting the same problem of low accuracy) You need to change metrics from 'accuracy' to 'binary_accuracy' to get the same values shown in the video.

Thanks for watching ❤️. You can comment below your doubts.

@@KGPTalkie Thanks for Your reply. i written coden on multilable image classification , working good, given right predict value when i used image from data/horse/horse.1.jpg , but when i tried to predict with one downloaded Horse image from Downloads it could n't be predict it ? why ? what was the issue . why it couldn't predict different location of image ? Please help me Code: from __future__ import generator_stop ​ import os,cv2 import numpy as np import matplotlib.pyplot as plt ​ from sklearn.utils import shuffle from sklearn.model_selection import train_test_split ​ from keras import backend as K K.set_image_dim_ordering('th') from keras.utils import np_utils from tensorflow.keras.models import Sequential,load_model from keras.layers import Activation ,Dropout,Flatten,Dense from keras.layers.convolutional import Convolution2D, MaxPooling2D from keras.utils import plot_model from keras.optimizers import SGD,RMSprop,adam ​ ​ ##---------------------- PATH = os.getcwd() # Define data path data_path = PATH + '/data' data_dir_list = os.listdir(data_path) ​ img_rows=128 img_cols=128 num_channel=1 num_epoch=20 ​ # Define the number of classes num_classes = 4 ​ img_data_list=[] ​ for dataset in data_dir_list: img_list=os.listdir(data_path+'/'+ dataset) print ('Loaded the images of dataset-'+'{} '.format(dataset)) for img in img_list: input_img=cv2.imread(data_path + '/'+ dataset + '/'+ img ) input_img=cv2.cvtColor(input_img, cv2.COLOR_BGR2GRAY) input_img_resize=cv2.resize(input_img,(128,128)) img_data_list.append(input_img_resize) ​ img_data = np.array(img_data_list) img_data = img_data.astype('float32') img_data /= 255 print (img_data.shape) ​ if num_channel==1: if K.image_dim_ordering()=='th': img_data= np.expand_dims(img_data, axis=1) print (img_data.shape) else: img_data= np.expand_dims(img_data, axis=4) print (img_data.shape) else: if K.image_dim_ordering()=='th': img_data=np.rollaxis(img_data,3,1) print (img_data.shape) ​ ​ ​ ##----------------------------------------------------------------------------------------------------------------- num_classes = 4 ​ num_of_samples = img_data.shape[0] labels = np.ones((num_of_samples,),dtype='int64') ​ labels[0:202]=0 labels[202:404]=1 labels[404:606]=2 labels[606:]=3 names = ['cats','dogs','horses','humans'] # convert class labels to on-hot encoding Y = np_utils.to_categorical(labels, num_classes) ​ #Shuffle the dataset x,y = shuffle(img_data,Y, random_state=2) # Split the dataset X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=2) input_shape=img_data[0].shape ##----------------------------------------------------------------------------------------------------------------- model = Sequential() ​ model.add(Convolution2D(32, 3,3,border_mode='same',input_shape=input_shape)) model.add(Activation('relu')) model.add(Convolution2D(32, 3, 3)) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.5)) ​ model.add(Convolution2D(64, 3, 3)) model.add(Activation('relu')) #model.add(Convolution2D(64, 3, 3)) #model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.5)) ​ model.add(Flatten()) model.add(Dense(64)) model.add(Activation('relu')) model.add(Dropout(0.5)) model.add(Dense(num_classes)) model.add(Activation('softmax')) ​ model.compile(loss='categorical_crossentropy', optimizer='rmsprop',metrics=["accuracy"]) model.summary() ​ hist = model.fit(X_train, y_train, batch_size=16, epochs=num_epoch, verbose=1, validation_data=(X_test, y_test)) ​ ########## TEST IMAGE ######################### test_image = cv2.imread('data1/dogs/dog.103.jpg') test_image=cv2.cvtColor(test_image, cv2.COLOR_BGR2GRAY) test_image=cv2.resize(test_image,(128,128)) test_image = np.array(test_image) test_image = test_image.astype('float32') test_image /= 255 print (test_image.shape) if num_channel==1: if K.image_dim_ordering()=='th': test_image= np.expand_dims(test_image, axis=0) test_image= np.expand_dims(test_image, axis=0) print (test_image.shape) else: test_image= np.expand_dims(test_image, axis=3) test_image= np.expand_dims(test_image, axis=0) print (test_image.shape) else: if K.image_dim_ordering()=='th': test_image=np.rollaxis(test_image,2,0) test_image= np.expand_dims(test_image, axis=0) print (test_image.shape) else: test_image= np.expand_dims(test_image, axis=0) print (test_image.shape) # Predicting the test image print((model.predict(test_image))) print(model.predict_classes(test_image))

Dense layers are often called flattening layers too. It means, no matter what is dimensions of your data, it will flat the layer. So use some hit and trail method to find optimal performance.

@@KGPTalkie thanks for your feedback. Let me put it this way: my labels (y_train) is not a vector, it is a 3D array. output_dimensions=y_train.shape= 1500x15x15 where 1500 is the number of inputs how do I implement it in the output layer? I tried (units=(15,15) in the dense layer, but it didn't work.

It needs lots of RAM. Once your notebook crashes then it will ask you to increase your RAM. You need to accept as Yes and increase RAM. It will work.

thanks. correction done.

@@KGPTalkie I love the way u explain keep posting videos . Google Colab is not giving any option to increase the RAM size with free of cost so there is any other alternate way. possible.

Hi, Google is not providing now free version upgrade. Reduce batchsize by half.

Yes can use one.

waiting for your positive reply.

Hi thanks for watching ❤️. I will make one once I come back from my leave.

@@KGPTalkie okk sir will be waiting for that video making a face recognition on multiple faces on a live feed using cnn model

@@KGPTalkie You are the best,please keep contributing! please start a series on OpenCV,a request from your Fan!

and it keep asking me to view runlogs

Hi, Google had recently changed it policy. Now they do not allow to increase the RAM. There is a pro version of colab but it is available only US. You might need to reduce your dataset and load it in chunks to reduce memory overhead.

@@KGPTalkie omg I see... Thank you for the reply!

@@KGPTalkie sorry for disturbing, can I have the coding of creating csv file?

also some with .png

Yes. It is a huge dataset.

Hi, Deep learning models needs lots of testing. I found 2 layers were performing better.

@@KGPTalkie Thank You

Hi Thanks for watching. This is multi class example TensorFlow 2.0 Tutorial for Beginners 15 - Malaria Parasite Detection Using CNN

@@KGPTalkie Hi Thanks a lot for the quick response. but isn't that considered a binary class classification ? i work with 10 classes. doesn't that affect the code ? do i need to do changes ? please help i'm really struggling. thanks a lot!

Yes you need to make some changes. Please cifar tutorial to make changes in the last layer.

reduce the batch size and then train. otherwise reduce input image size .

Then I can't say much about Pro Version. But in normal CoLab it comes everytime. Initially RAM size was 12GB later it gets 25GB. Check your RAM size.

Experiencing the same. By any chance, have you already discovered why? :)

@@dave.jammin Same here do you know why ?

Thanks for watching it. you need to share more info.

Experiencing the same. By any chance, have you already discovered why? :)

Same here. Both accuracy and val_accuracy are extremely low.

I have the same problem, do you have a solution ?

The default accuracy calculation has changed. Now it gives more realistic values. Previously result giving 0 to all would have been "great" as for most of the values are 0 in all of the movies. You can get the "same" result by manually changing metrics.