cifar_10.py

# -*- coding: utf-8 -*-
"""CIFAR-10.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1gz2WNaAci8BVhAa4ON5xSsjZwzi7zIp_
"""

import tensorflow as tf
import numpy as np

def load_dataset(path,batchid):
  with open(path + '/data_batch_' + str(batchid), mode='rb') as file:
    batch=pickle.load(file,encoding='latin')
    features=batch['data'].reshape((len(batch['data']),3,32,32)).transpose(0,2,3,1)
    labels=batch['label']
    return features,labels

def load_names():
  return ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']

def normalise(x):
  min_val=np.min(x)
  max_val=np.max(x)
  x=(x-min_val)/(max_val-min_val)

def one_hot_encoding(x):
  encoded=np.zeroes((len(x),10))a

def preprocess_save(normalise,one_hot_encoding,features,labels,file):
  features=normalise(features)
  labels=one_hot_encoding(labels)
  pickle.dump((features,labels),open(filename,'wb'))

def preprocess_save_data(path,normalise,one_hot_encoding):
  n_batches=5
  valid_features=[]
  valid_labels=[]
  for batch_i in range(1,n_batches+1):