In [2]:
# %load /home/sjkim/.jupyter/head.py
%matplotlib inline
%load_ext autoreload
%autoreload 2
from importlib import reload
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import os
#os.environ["CUDA_VISIBLE_DEVICES"]="0"
# seaborn
#import seaborn as sns
#sns.set( style = 'white', font_scale = 1.7)
#sns.set_style('ticks')
#plt.rcParams['savefig.dpi'] = 200
# font for matplotlib
#import matplotlib
#import matplotlib.font_manager as fm
#fm.get_fontconfig_fonts()
#font_location = '/usr/share/fonts/truetype/nanum/NanumGothicBold.ttf'
#font_name = fm.FontProperties(fname=font_location).get_name()
#matplotlib.rc('font', family=font_name)
The autoreload extension is already loaded. To reload it, use:
%reload_ext autoreload
In [2]:
import ex4_2_cnn_cifar10_cl as example
Using TensorFlow backend.
In [3]:
example.main()
(40000, 32, 32, 3) (40000, 1)
X_train shape: (40000, 32, 32, 3)
40000 train samples
10000 test samples
data.input_shape (32, 32, 3)
Train on 40000 samples, validate on 10000 samples
Epoch 1/100
40000/40000 [==============================] - 5s - loss: 1.8609 - acc: 0.3241 - val_loss: 1.4764 - val_acc: 0.4754
Epoch 2/100
40000/40000 [==============================] - 3s - loss: 1.4631 - acc: 0.4799 - val_loss: 1.2612 - val_acc: 0.5520
Epoch 3/100
40000/40000 [==============================] - 3s - loss: 1.2995 - acc: 0.5408 - val_loss: 1.1802 - val_acc: 0.5872
Epoch 4/100
40000/40000 [==============================] - 3s - loss: 1.2007 - acc: 0.5781 - val_loss: 1.0836 - val_acc: 0.6220
Epoch 5/100
40000/40000 [==============================] - 3s - loss: 1.1309 - acc: 0.6028 - val_loss: 1.0639 - val_acc: 0.6287
Epoch 6/100
40000/40000 [==============================] - 3s - loss: 1.0644 - acc: 0.6272 - val_loss: 1.0287 - val_acc: 0.6385
Epoch 7/100
40000/40000 [==============================] - 3s - loss: 1.0146 - acc: 0.6456 - val_loss: 0.9954 - val_acc: 0.6490
Epoch 8/100
40000/40000 [==============================] - 3s - loss: 0.9783 - acc: 0.6562 - val_loss: 0.9608 - val_acc: 0.6596
Epoch 9/100
40000/40000 [==============================] - 3s - loss: 0.9396 - acc: 0.6723 - val_loss: 0.9566 - val_acc: 0.6680
Epoch 10/100
40000/40000 [==============================] - 3s - loss: 0.9059 - acc: 0.6828 - val_loss: 0.9368 - val_acc: 0.6720
Epoch 11/100
40000/40000 [==============================] - 3s - loss: 0.8743 - acc: 0.6954 - val_loss: 0.9030 - val_acc: 0.6842
Epoch 12/100
40000/40000 [==============================] - 3s - loss: 0.8435 - acc: 0.7056 - val_loss: 0.9098 - val_acc: 0.6802
Epoch 13/100
40000/40000 [==============================] - 3s - loss: 0.8148 - acc: 0.7138 - val_loss: 0.9176 - val_acc: 0.6825
Epoch 14/100
40000/40000 [==============================] - 3s - loss: 0.7849 - acc: 0.7253 - val_loss: 0.9149 - val_acc: 0.6807
Epoch 15/100
40000/40000 [==============================] - 3s - loss: 0.7619 - acc: 0.7336 - val_loss: 0.8915 - val_acc: 0.6891
Epoch 16/100
40000/40000 [==============================] - 3s - loss: 0.7349 - acc: 0.7426 - val_loss: 0.8787 - val_acc: 0.6967
Epoch 17/100
40000/40000 [==============================] - 3s - loss: 0.7140 - acc: 0.7503 - val_loss: 0.8941 - val_acc: 0.6939
Epoch 18/100
40000/40000 [==============================] - 3s - loss: 0.6896 - acc: 0.7608 - val_loss: 0.9127 - val_acc: 0.6846
Epoch 19/100
40000/40000 [==============================] - 3s - loss: 0.6714 - acc: 0.7645 - val_loss: 0.8656 - val_acc: 0.7020
Epoch 20/100
40000/40000 [==============================] - 3s - loss: 0.6497 - acc: 0.7724 - val_loss: 0.8803 - val_acc: 0.6970
Epoch 21/100
40000/40000 [==============================] - 3s - loss: 0.6332 - acc: 0.7774 - val_loss: 0.8810 - val_acc: 0.7061
Epoch 22/100
40000/40000 [==============================] - 3s - loss: 0.6158 - acc: 0.7830 - val_loss: 0.8592 - val_acc: 0.7020
Epoch 23/100
40000/40000 [==============================] - 3s - loss: 0.5907 - acc: 0.7962 - val_loss: 0.8724 - val_acc: 0.7035
Epoch 24/100
40000/40000 [==============================] - 3s - loss: 0.5782 - acc: 0.7969 - val_loss: 0.8659 - val_acc: 0.7088
Epoch 25/100
40000/40000 [==============================] - 3s - loss: 0.5647 - acc: 0.7988 - val_loss: 0.8644 - val_acc: 0.7095
Epoch 26/100
40000/40000 [==============================] - 3s - loss: 0.5483 - acc: 0.8075 - val_loss: 0.8671 - val_acc: 0.7073
Epoch 27/100
40000/40000 [==============================] - 3s - loss: 0.5284 - acc: 0.8141 - val_loss: 0.8595 - val_acc: 0.7090
Epoch 28/100
40000/40000 [==============================] - 3s - loss: 0.5209 - acc: 0.8160 - val_loss: 0.8777 - val_acc: 0.7115
Epoch 29/100
40000/40000 [==============================] - 3s - loss: 0.5057 - acc: 0.8218 - val_loss: 0.9017 - val_acc: 0.7155
Epoch 30/100
40000/40000 [==============================] - 3s - loss: 0.4939 - acc: 0.8287 - val_loss: 0.9373 - val_acc: 0.7044
Epoch 31/100
40000/40000 [==============================] - 3s - loss: 0.4838 - acc: 0.8313 - val_loss: 0.8926 - val_acc: 0.7079
Epoch 32/100
40000/40000 [==============================] - 3s - loss: 0.4728 - acc: 0.8336 - val_loss: 0.9065 - val_acc: 0.7123
Epoch 33/100
40000/40000 [==============================] - 3s - loss: 0.4591 - acc: 0.8376 - val_loss: 0.8891 - val_acc: 0.7169
Epoch 34/100
40000/40000 [==============================] - 3s - loss: 0.4541 - acc: 0.8418 - val_loss: 0.9681 - val_acc: 0.7144
Epoch 35/100
40000/40000 [==============================] - 3s - loss: 0.4441 - acc: 0.8438 - val_loss: 0.9162 - val_acc: 0.7131
Epoch 36/100
40000/40000 [==============================] - 3s - loss: 0.4319 - acc: 0.8484 - val_loss: 0.9611 - val_acc: 0.7159
Epoch 37/100
40000/40000 [==============================] - 3s - loss: 0.4221 - acc: 0.8517 - val_loss: 0.9162 - val_acc: 0.7188
Epoch 38/100
40000/40000 [==============================] - 3s - loss: 0.4211 - acc: 0.8541 - val_loss: 0.9267 - val_acc: 0.7127
Epoch 39/100
40000/40000 [==============================] - 3s - loss: 0.4131 - acc: 0.8548 - val_loss: 0.8874 - val_acc: 0.7113
Epoch 40/100
40000/40000 [==============================] - 3s - loss: 0.4054 - acc: 0.8579 - val_loss: 0.9146 - val_acc: 0.7174
Epoch 41/100
40000/40000 [==============================] - 3s - loss: 0.3998 - acc: 0.8575 - val_loss: 0.8720 - val_acc: 0.7156
Epoch 42/100
40000/40000 [==============================] - 3s - loss: 0.3945 - acc: 0.8613 - val_loss: 0.9557 - val_acc: 0.7219
Epoch 43/100
40000/40000 [==============================] - 3s - loss: 0.3846 - acc: 0.8659 - val_loss: 0.8766 - val_acc: 0.7133
Epoch 44/100
40000/40000 [==============================] - 3s - loss: 0.3799 - acc: 0.8668 - val_loss: 0.9342 - val_acc: 0.7180
Epoch 45/100
40000/40000 [==============================] - 3s - loss: 0.3772 - acc: 0.8672 - val_loss: 0.9253 - val_acc: 0.7101
Epoch 46/100
40000/40000 [==============================] - 3s - loss: 0.3776 - acc: 0.8696 - val_loss: 0.9656 - val_acc: 0.7167
Epoch 47/100
40000/40000 [==============================] - 3s - loss: 0.3687 - acc: 0.8732 - val_loss: 0.9579 - val_acc: 0.7200
Epoch 48/100
40000/40000 [==============================] - 3s - loss: 0.3741 - acc: 0.8701 - val_loss: 1.0333 - val_acc: 0.7170
Epoch 49/100
40000/40000 [==============================] - 3s - loss: 0.3592 - acc: 0.8744 - val_loss: 0.9238 - val_acc: 0.6980
Epoch 50/100
40000/40000 [==============================] - 3s - loss: 0.3634 - acc: 0.8726 - val_loss: 1.0218 - val_acc: 0.7169
Epoch 51/100
40000/40000 [==============================] - 3s - loss: 0.3619 - acc: 0.8743 - val_loss: 0.9055 - val_acc: 0.7113
Epoch 52/100
40000/40000 [==============================] - 3s - loss: 0.3527 - acc: 0.8779 - val_loss: 0.9336 - val_acc: 0.7150
Epoch 53/100
40000/40000 [==============================] - 3s - loss: 0.3500 - acc: 0.8783 - val_loss: 1.1906 - val_acc: 0.7207
Epoch 54/100
40000/40000 [==============================] - 3s - loss: 0.3468 - acc: 0.8807 - val_loss: 1.0368 - val_acc: 0.7222
Epoch 55/100
40000/40000 [==============================] - 3s - loss: 0.3518 - acc: 0.8787 - val_loss: 1.0134 - val_acc: 0.7224
Epoch 56/100
40000/40000 [==============================] - 3s - loss: 0.3424 - acc: 0.8828 - val_loss: 1.0886 - val_acc: 0.7198
Epoch 57/100
40000/40000 [==============================] - 3s - loss: 0.3452 - acc: 0.8809 - val_loss: 0.9535 - val_acc: 0.7162
Epoch 58/100
40000/40000 [==============================] - 3s - loss: 0.3527 - acc: 0.8800 - val_loss: 1.1648 - val_acc: 0.7177
Epoch 59/100
40000/40000 [==============================] - 3s - loss: 0.3363 - acc: 0.8865 - val_loss: 1.0878 - val_acc: 0.7166
Epoch 60/100
40000/40000 [==============================] - 3s - loss: 0.3419 - acc: 0.8837 - val_loss: 0.9495 - val_acc: 0.6870
Epoch 61/100
40000/40000 [==============================] - 3s - loss: 0.3468 - acc: 0.8833 - val_loss: 0.9590 - val_acc: 0.7177
Epoch 62/100
40000/40000 [==============================] - 3s - loss: 0.3481 - acc: 0.8828 - val_loss: 1.0042 - val_acc: 0.7162
Epoch 63/100
40000/40000 [==============================] - 3s - loss: 0.3392 - acc: 0.8861 - val_loss: 0.8929 - val_acc: 0.7106
Epoch 64/100
40000/40000 [==============================] - 3s - loss: 0.3473 - acc: 0.8833 - val_loss: 1.1903 - val_acc: 0.7171
Epoch 65/100
40000/40000 [==============================] - 3s - loss: 0.3456 - acc: 0.8835 - val_loss: 1.0837 - val_acc: 0.7138
Epoch 66/100
40000/40000 [==============================] - 3s - loss: 0.3519 - acc: 0.8841 - val_loss: 0.9791 - val_acc: 0.7198
Epoch 67/100
40000/40000 [==============================] - 3s - loss: 0.3387 - acc: 0.8879 - val_loss: 1.0049 - val_acc: 0.7189
Epoch 68/100
40000/40000 [==============================] - 3s - loss: 0.3372 - acc: 0.8857 - val_loss: 0.9725 - val_acc: 0.7130
Epoch 69/100
40000/40000 [==============================] - 3s - loss: 0.3405 - acc: 0.8841 - val_loss: 1.0522 - val_acc: 0.7215
Epoch 70/100
40000/40000 [==============================] - 3s - loss: 0.3431 - acc: 0.8852 - val_loss: 1.0734 - val_acc: 0.7172
Epoch 71/100
40000/40000 [==============================] - 3s - loss: 0.3403 - acc: 0.8870 - val_loss: 0.9666 - val_acc: 0.7182
Epoch 72/100
40000/40000 [==============================] - 3s - loss: 0.3424 - acc: 0.8885 - val_loss: 0.9276 - val_acc: 0.7158
Epoch 73/100
40000/40000 [==============================] - 3s - loss: 0.3381 - acc: 0.8867 - val_loss: 0.9366 - val_acc: 0.7100
Epoch 74/100
40000/40000 [==============================] - 3s - loss: 0.3456 - acc: 0.8843 - val_loss: 0.9476 - val_acc: 0.6823
Epoch 75/100
40000/40000 [==============================] - 3s - loss: 0.3415 - acc: 0.8873 - val_loss: 1.0602 - val_acc: 0.7183
Epoch 76/100
40000/40000 [==============================] - 3s - loss: 0.3390 - acc: 0.8864 - val_loss: 1.0361 - val_acc: 0.7162
Epoch 77/100
40000/40000 [==============================] - 3s - loss: 0.3414 - acc: 0.8872 - val_loss: 0.8997 - val_acc: 0.7015
Epoch 78/100
40000/40000 [==============================] - 3s - loss: 0.3387 - acc: 0.8868 - val_loss: 1.4872 - val_acc: 0.7158
Epoch 79/100
40000/40000 [==============================] - 3s - loss: 0.3470 - acc: 0.8850 - val_loss: 1.0980 - val_acc: 0.7172
Epoch 80/100
40000/40000 [==============================] - 3s - loss: 0.3444 - acc: 0.8871 - val_loss: 1.3997 - val_acc: 0.7186
Epoch 81/100
40000/40000 [==============================] - 3s - loss: 0.3380 - acc: 0.8870 - val_loss: 1.1492 - val_acc: 0.7220
Epoch 82/100
40000/40000 [==============================] - 3s - loss: 0.3506 - acc: 0.8854 - val_loss: 0.9545 - val_acc: 0.7144
Epoch 83/100
40000/40000 [==============================] - 3s - loss: 0.3376 - acc: 0.8872 - val_loss: 1.3168 - val_acc: 0.7172
Epoch 84/100
40000/40000 [==============================] - 3s - loss: 0.3424 - acc: 0.8891 - val_loss: 0.9385 - val_acc: 0.6858
Epoch 85/100
40000/40000 [==============================] - 3s - loss: 0.3560 - acc: 0.8835 - val_loss: 1.1772 - val_acc: 0.7217
Epoch 86/100
40000/40000 [==============================] - 3s - loss: 0.3395 - acc: 0.8889 - val_loss: 1.1543 - val_acc: 0.7195
Epoch 87/100
40000/40000 [==============================] - 3s - loss: 0.3561 - acc: 0.8852 - val_loss: 0.9171 - val_acc: 0.7036
Epoch 88/100
40000/40000 [==============================] - 3s - loss: 0.3527 - acc: 0.8854 - val_loss: 1.0270 - val_acc: 0.7161
Epoch 89/100
40000/40000 [==============================] - 3s - loss: 0.3530 - acc: 0.8863 - val_loss: 0.9804 - val_acc: 0.6845
Epoch 90/100
40000/40000 [==============================] - 3s - loss: 0.3512 - acc: 0.8868 - val_loss: 1.1917 - val_acc: 0.7179
Epoch 91/100
40000/40000 [==============================] - 3s - loss: 0.3406 - acc: 0.8883 - val_loss: 1.0429 - val_acc: 0.7221
Epoch 92/100
40000/40000 [==============================] - 3s - loss: 0.3470 - acc: 0.8867 - val_loss: 1.0645 - val_acc: 0.7222
Epoch 93/100
40000/40000 [==============================] - 3s - loss: 0.3573 - acc: 0.8845 - val_loss: 0.9055 - val_acc: 0.7083
Epoch 94/100
40000/40000 [==============================] - 3s - loss: 0.3519 - acc: 0.8866 - val_loss: 1.4045 - val_acc: 0.7231
Epoch 95/100
40000/40000 [==============================] - 3s - loss: 0.3598 - acc: 0.8854 - val_loss: 1.3076 - val_acc: 0.7233
Epoch 96/100
40000/40000 [==============================] - 3s - loss: 0.3448 - acc: 0.8892 - val_loss: 0.9616 - val_acc: 0.7156
Epoch 97/100
40000/40000 [==============================] - 3s - loss: 0.3658 - acc: 0.8829 - val_loss: 0.9370 - val_acc: 0.7166
Epoch 98/100
40000/40000 [==============================] - 3s - loss: 0.3596 - acc: 0.8827 - val_loss: 1.0914 - val_acc: 0.7212
Epoch 99/100
40000/40000 [==============================] - 3s - loss: 0.3552 - acc: 0.8844 - val_loss: 0.9374 - val_acc: 0.7076
Epoch 100/100
40000/40000 [==============================] - 3s - loss: 0.3605 - acc: 0.8839 - val_loss: 1.3648 - val_acc: 0.7220
Confusion matrix
[[767 21 60 13 21 11 11 9 64 27]
[ 14 838 15 8 6 5 18 2 20 60]
[ 47 5 674 43 85 73 48 23 12 6]
[ 15 5 91 509 69 216 84 22 13 6]
[ 26 0 78 44 670 42 45 47 10 5]
[ 3 3 88 159 33 653 36 33 4 4]
[ 4 3 55 57 33 18 775 2 4 2]
[ 14 3 41 39 67 75 5 741 5 9]
[ 65 40 19 19 6 8 8 3 797 24]
[ 38 118 8 20 6 10 13 16 15 796]]
Test score: 1.3648323123
Test accuracy: 0.722
Output results are saved in output_b859dff0-a53c-42a4-b077-ba748151f9ec
Content source: jskDr/keraspp
Similar notebooks: