- Basic Information
- Official website: http://www.cs.toronto.edu/~kriz/cifar.html
- Total 60000 Pictures: 50000 for training and 10,000 for testing
- Image size: 32X32
- Image data set is divided into 10 categories: 6000 per class
- Data set directory structure after downloading codecs:
- Read, print and save images of the specified data set:
import pickle import matplotlib.pyplot as plt CIFAR_DIR = " cifar10_data / CIFAR-10-bin-Batches / data_batch_1.bin " # dataset path with Open (CIFAR_DIR, ' RB ' ) AS F: data = pickle.load(f, encoding='bytes') Print ( ' ---------- batch1 the basic information ------------- ' ) Print ( ' Data Type Data: ' , type (Data)) # Output < class' dict '> Print ( ' dictionary key name: ' , data.keys ()) # output dict_keys ([b'filenames', b'data', b'labels', b'batch_label ']) Print ( ' bdata datatype ' , type (data [B ' data ' ])) # output <class' numpy.ndarray'> Print ( ' bdata shape data ' ,data[b'data'] .shape) # output (10000, 3072) Description of 10,000 samples, 3,072 feature index =. 4 # print first few images Print ( ' -----------% d of Image - --------- ' % index) Print ( ' of the filenames: ' , Data [B ' of the filenames ' ] [index]) Print ( ' Labels: ' , Data [B ' Labels ' ] [index]) Print ( ' batch_label: ' , Data [B ' batch_label ' ] [index]) image_arr = Data [B ' Data ' ] [index] # Take the first sample index image_arr image_arr.reshape = ((. 3, 32, 32)) # will change the shape of one-dimensional vectors to obtain a high :( such a tuple, width , channels) image_arr image_arr.transpose = ((. 1, 2 , 0)) plt.imshow (image_arr) # output picture plt.savefig ( " cifar10_data / RAW /% d.png " % index) # save the image plt.show ()
- Print out pictures
-
Tensorflow machine learning portal --cifar10 data set to read, display and save
Guess you like
Origin www.cnblogs.com/Fengqiao/p/cifar10_read.html
Recommended
Ranking