keras_image_classification_multi_categorical_crossentropy.log

# Create output layer with 4 node(4-class) and sigmoid activation
#output = layers.Dense(4, activation='sigmoid')(x)
output = layers.Dense(4, activation='softmax')(x)

# Create model:
model = Model(img_input, output)

# Compile model
model.compile(loss='categorical_crossentropy',
              optimizer=Adam(),#RMSprop(lr=0.001),
              #,optimizer=SGD(lr=1e-4, momentum=0.9)
              metrics=['acc'])

=============train_datagen & val_datagen batch_size=14 & steps_per_epoch=95/14
7/6 - 3s - loss: 1.4740 - acc: 0.3053 - val_loss: 1.3883 - val_acc: 0.3333
Epoch 2/50
7/6 - 1s - loss: 1.4262 - acc: 0.2211 - val_loss: 1.3878 - val_acc: 0.2222
Epoch 3/50
7/6 - 1s - loss: 1.3818 - acc: 0.2737 - val_loss: 1.3833 - val_acc: 0.3333
Epoch 4/50
7/6 - 1s - loss: 1.3863 - acc: 0.2632 - val_loss: 1.3811 - val_acc: 0.3333
Epoch 5/50
7/6 - 1s - loss: 1.3620 - acc: 0.3579 - val_loss: 1.3793 - val_acc: 0.3333
Epoch 6/50
7/6 - 1s - loss: 1.3533 - acc: 0.2211 - val_loss: 1.3701 - val_acc: 0.2778
Epoch 7/50
7/6 - 1s - loss: 1.3411 - acc: 0.3158 - val_loss: 1.3575 - val_acc: 0.3333
Epoch 8/50
7/6 - 1s - loss: 1.3088 - acc: 0.3474 - val_loss: 1.3376 - val_acc: 0.5000
Epoch 9/50
7/6 - 1s - loss: 1.2142 - acc: 0.5579 - val_loss: 1.2307 - val_acc: 0.6111
Epoch 10/50
7/6 - 1s - loss: 1.0678 - acc: 0.5053 - val_loss: 1.0566 - val_acc: 0.7222
Epoch 11/50
7/6 - 1s - loss: 0.9914 - acc: 0.5579 - val_loss: 1.1081 - val_acc: 0.3889
Epoch 12/50
7/6 - 1s - loss: 0.9393 - acc: 0.5684 - val_loss: 0.9826 - val_acc: 0.6111
Epoch 13/50
7/6 - 1s - loss: 0.6940 - acc: 0.7053 - val_loss: 0.8639 - val_acc: 0.6111
Epoch 14/50
7/6 - 1s - loss: 0.8318 - acc: 0.6000 - val_loss: 0.7978 - val_acc: 0.5556
Epoch 15/50
7/6 - 1s - loss: 0.5273 - acc: 0.7579 - val_loss: 0.7448 - val_acc: 0.6111
Epoch 16/50
7/6 - 1s - loss: 0.6337 - acc: 0.7158 - val_loss: 0.8857 - val_acc: 0.6111
Epoch 17/50
7/6 - 1s - loss: 0.5243 - acc: 0.7684 - val_loss: 0.6613 - val_acc: 0.8333
Epoch 18/50
7/6 - 1s - loss: 0.3622 - acc: 0.8316 - val_loss: 0.9130 - val_acc: 0.7778
Epoch 19/50
7/6 - 1s - loss: 0.3801 - acc: 0.8421 - val_loss: 0.7626 - val_acc: 0.8333
Epoch 20/50
7/6 - 1s - loss: 0.2820 - acc: 0.8737 - val_loss: 0.7489 - val_acc: 0.7778
Epoch 21/50
7/6 - 1s - loss: 0.3044 - acc: 0.8737 - val_loss: 1.1160 - val_acc: 0.5000
Epoch 22/50
7/6 - 1s - loss: 0.2443 - acc: 0.9263 - val_loss: 0.7759 - val_acc: 0.7778
Epoch 23/50
7/6 - 1s - loss: 0.2240 - acc: 0.9474 - val_loss: 1.0746 - val_acc: 0.5556
Epoch 24/50
7/6 - 1s - loss: 0.3261 - acc: 0.8737 - val_loss: 0.6705 - val_acc: 0.7778
Epoch 25/50
7/6 - 1s - loss: 0.1911 - acc: 0.9053 - val_loss: 0.9043 - val_acc: 0.6667
Epoch 26/50
7/6 - 1s - loss: 0.1906 - acc: 0.9263 - val_loss: 0.9787 - val_acc: 0.7778
Epoch 27/50
7/6 - 1s - loss: 0.1434 - acc: 0.9684 - val_loss: 0.8488 - val_acc: 0.6667

=-====================== best model save(17/50)
[True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
20
24
0.8333333333333334
======================= end save

=============train_datagen & val_datagen batch_size=8 & steps_per_epoch=95/8

Epoch 2/50
12/11 - 1s - loss: 1.3570 - acc: 0.2632 - val_loss: 1.3676 - val_acc: 0.2222
Epoch 3/50
12/11 - 1s - loss: 1.3294 - acc: 0.2632 - val_loss: 1.3248 - val_acc: 0.2222
Epoch 4/50
12/11 - 1s - loss: 1.2341 - acc: 0.4211 - val_loss: 1.2516 - val_acc: 0.5000
Epoch 5/50
12/11 - 1s - loss: 1.1012 - acc: 0.5368 - val_loss: 1.1262 - val_acc: 0.3333
Epoch 6/50
12/11 - 1s - loss: 0.9315 - acc: 0.5579 - val_loss: 0.9974 - val_acc: 0.4444
Epoch 7/50
12/11 - 1s - loss: 0.7087 - acc: 0.7474 - val_loss: 0.7988 - val_acc: 0.7778
Epoch 8/50
12/11 - 1s - loss: 0.5658 - acc: 0.7895 - val_loss: 0.8148 - val_acc: 0.7778
Epoch 9/50
12/11 - 1s - loss: 0.4630 - acc: 0.8105 - val_loss: 0.6758 - val_acc: 0.7778
Epoch 10/50
12/11 - 1s - loss: 0.2281 - acc: 0.9263 - val_loss: 0.5640 - val_acc: 0.9444
Epoch 11/50
12/11 - 1s - loss: 0.2970 - acc: 0.8947 - val_loss: 0.7121 - val_acc: 0.8333
Epoch 12/50
12/11 - 1s - loss: 0.3239 - acc: 0.8421 - val_loss: 0.5827 - val_acc: 0.8889
Epoch 13/50
12/11 - 1s - loss: 0.1958 - acc: 0.9263 - val_loss: 0.4920 - val_acc: 0.8889
Epoch 14/50
12/11 - 1s - loss: 0.1765 - acc: 0.9474 - val_loss: 0.4403 - val_acc: 0.9444
Epoch 15/50
12/11 - 1s - loss: 0.1713 - acc: 0.9263 - val_loss: 0.6751 - val_acc: 0.8889
Epoch 16/50
12/11 - 1s - loss: 0.1693 - acc: 0.9474 - val_loss: 0.3930 - val_acc: 0.9444
Epoch 17/50
12/11 - 1s - loss: 0.1037 - acc: 0.9789 - val_loss: 0.6398 - val_acc: 0.8889
Epoch 18/50
12/11 - 1s - loss: 0.0902 - acc: 0.9579 - val_loss: 0.5238 - val_acc: 0.9444
Epoch 19/50
12/11 - 1s - loss: 0.0667 - acc: 0.9579 - val_loss: 0.6423 - val_acc: 0.9444
Epoch 20/50
12/11 - 1s - loss: 0.0278 - acc: 0.9895 - val_loss: 0.7196 - val_acc: 0.8889
Epoch 21/50
12/11 - 1s - loss: 0.0612 - acc: 0.9684 - val_loss: 0.6409 - val_acc: 0.9444
Epoch 22/50
12/11 - 1s - loss: 0.0379 - acc: 0.9895 - val_loss: 1.1552 - val_acc: 0.8889
Epoch 23/50
12/11 - 1s - loss: 0.0626 - acc: 0.9895 - val_loss: 0.8743 - val_acc: 0.9444
Epoch 24/50
12/11 - 1s - loss: 0.0969 - acc: 0.9579 - val_loss: 0.4726 - val_acc: 0.9444
Epoch 25/50
12/11 - 1s - loss: 0.0809 - acc: 0.9789 - val_loss: 0.7113 - val_acc: 0.9444
Epoch 26/50
12/11 - 1s - loss: 0.0730 - acc: 0.9789 - val_loss: 0.6977 - val_acc: 0.9444

=-====================== best model save(17/50)
[True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
22
24
0.9166666666666666
======================= end save



=============train_datagen & val_datagen batch_size=6
2/12 - 1s - loss: 0.3249 - acc: 0.8310 - val_loss: 0.4781 - val_acc: 0.7778
Epoch 20/50
12/12 - 1s - loss: 0.5434 - acc: 0.7606 - val_loss: 0.6544 - val_acc: 0.6111
Epoch 21/50
12/12 - 1s - loss: 0.4161 - acc: 0.8732 - val_loss: 0.4632 - val_acc: 0.7778
Epoch 22/50
12/12 - 1s - loss: 0.6663 - acc: 0.8169 - val_loss: 0.6630 - val_acc: 0.6667
Epoch 23/50
12/12 - 1s - loss: 0.3692 - acc: 0.8333 - val_loss: 0.5308 - val_acc: 0.8333
Epoch 24/50
12/12 - 1s - loss: 0.2377 - acc: 0.9155 - val_loss: 0.3685 - val_acc: 0.8889
Epoch 25/50
12/12 - 1s - loss: 0.1982 - acc: 0.9444 - val_loss: 0.1220 - val_acc: 0.9444
Epoch 26/50
12/12 - 1s - loss: 0.2708 - acc: 0.9143 - val_loss: 0.1988 - val_acc: 0.9444
Epoch 27/50
12/12 - 1s - loss: 0.2910 - acc: 0.8889 - val_loss: 0.5431 - val_acc: 0.7222
Epoch 28/50
12/12 - 1s - loss: 0.3683 - acc: 0.8169 - val_loss: 0.4691 - val_acc: 0.8889
Epoch 29/50
12/12 - 1s - loss: 0.2462 - acc: 0.9155 - val_loss: 0.3278 - val_acc: 0.9444
Epoch 30/50
12/12 - 1s - loss: 0.2702 - acc: 0.9028 - val_loss: 0.2364 - val_acc: 0.9444
Epoch 31/50
12/12 - 1s - loss: 0.0879 - acc: 0.9571 - val_loss: 0.1550 - val_acc: 0.9444
Epoch 32/50
12/12 - 1s - loss: 0.0536 - acc: 1.0000 - val_loss: 0.0984 - val_acc: 1.0000
Epoch 33/50
12/12 - 1s - loss: 0.0685 - acc: 0.9859 - val_loss: 0.2587 - val_acc: 0.9444
Epoch 34/50
12/12 - 1s - loss: 0.1105 - acc: 0.9861 - val_loss: 0.0693 - val_acc: 1.0000
Epoch 35/50
12/12 - 1s - loss: 0.0305 - acc: 0.9859 - val_loss: 0.1272 - val_acc: 0.9444
Epoch 36/50
12/12 - 1s - loss: 0.1022 - acc: 0.9437 - val_loss: 0.1554 - val_acc: 0.9444
Epoch 37/50
12/12 - 1s - loss: 0.1684 - acc: 0.9296 - val_loss: 0.4023 - val_acc: 0.8889
Epoch 38/50
12/12 - 1s - loss: 0.1232 - acc: 0.9722 - val_loss: 0.5180 - val_acc: 0.8333
Epoch 39/50
12/12 - 1s - loss: 0.0507 - acc: 0.9859 - val_loss: 0.3123 - val_acc: 0.8889
Epoch 40/50
12/12 - 1s - loss: 0.0076 - acc: 1.0000 - val_loss: 0.2028 - val_acc: 0.9444
Epoch 41/50
12/12 - 1s - loss: 0.0153 - acc: 0.9861 - val_loss: 0.1347 - val_acc: 0.9444
Epoch 42/50
12/12 - 1s - loss: 0.1496 - acc: 0.9437 - val_loss: 0.2540 - val_acc: 0.9444
Epoch 43/50
12/12 - 1s - loss: 0.0925 - acc: 0.9577 - val_loss: 0.2429 - val_acc: 0.9444
Epoch 44/50
12/12 - 1s - loss: 0.0257 - acc: 0.9859 - val_loss: 0.3292 - val_acc: 0.9444

=-====================== best model save(17/50)

22
24
0.9166666666666666
======================= end save

21
24
0.875

=============train_datagen & val_datagen batch_size=6
Epoch 1/50

  "Palette images with Transparency expressed in bytes should be "
12/12 - 3s - loss: 1.7387 - acc: 0.2394 - val_loss: 1.3862 - val_acc: 0.2222
Epoch 2/50
12/12 - 1s - loss: 1.3926 - acc: 0.2778 - val_loss: 1.3887 - val_acc: 0.2222
Epoch 3/50
12/12 - 1s - loss: 1.3959 - acc: 0.1857 - val_loss: 1.3877 - val_acc: 0.2222
Epoch 4/50
12/12 - 1s - loss: 1.3849 - acc: 0.2778 - val_loss: 1.3887 - val_acc: 0.2222
Epoch 5/50
12/12 - 1s - loss: 1.3857 - acc: 0.2676 - val_loss: 1.3898 - val_acc: 0.1667
Epoch 6/50
12/12 - 1s - loss: 1.3883 - acc: 0.2222 - val_loss: 1.3888 - val_acc: 0.2222
Epoch 7/50
12/12 - 1s - loss: 1.3847 - acc: 0.3286 - val_loss: 1.3889 - val_acc: 0.2222
Epoch 8/50
12/12 - 1s - loss: 1.3868 - acc: 0.2361 - val_loss: 1.3883 - val_acc: 0.2222
Epoch 9/50
12/12 - 1s - loss: 1.3816 - acc: 0.2817 - val_loss: 1.3891 - val_acc: 0.2222
Epoch 10/50
12/12 - 1s - loss: 1.3789 - acc: 0.3472 - val_loss: 1.3851 - val_acc: 0.2222
Epoch 11/50
12/12 - 1s - loss: 1.3962 - acc: 0.4000 - val_loss: 1.3881 - val_acc: 0.2222
Epoch 12/50
12/12 - 1s - loss: 1.3750 - acc: 0.3194 - val_loss: 1.3770 - val_acc: 0.5000
Epoch 13/50
12/12 - 1s - loss: 1.3784 - acc: 0.2958 - val_loss: 1.3708 - val_acc: 0.2778
Epoch 14/50
12/12 - 1s - loss: 1.3432 - acc: 0.3889 - val_loss: 1.3365 - val_acc: 0.3889
Epoch 15/50
12/12 - 1s - loss: 1.2499 - acc: 0.4000 - val_loss: 1.2181 - val_acc: 0.6667
Epoch 16/50
12/12 - 1s - loss: 1.2217 - acc: 0.4167 - val_loss: 1.3002 - val_acc: 0.3333
Epoch 17/50
12/12 - 1s - loss: 0.9832 - acc: 0.5694 - val_loss: 1.0487 - val_acc: 0.5556
Epoch 18/50
12/12 - 1s - loss: 0.9038 - acc: 0.5429 - val_loss: 0.7331 - val_acc: 0.7222
Epoch 19/50
12/12 - 1s - loss: 0.8739 - acc: 0.6389 - val_loss: 0.8316 - val_acc: 0.6667
Epoch 20/50
12/12 - 1s - loss: 0.6511 - acc: 0.6620 - val_loss: 0.6160 - val_acc: 0.8333
Epoch 21/50
12/12 - 1s - loss: 0.5675 - acc: 0.8333 - val_loss: 0.6635 - val_acc: 0.7222
Epoch 22/50
12/12 - 1s - loss: 0.9444 - acc: 0.6143 - val_loss: 0.9749 - val_acc: 0.5000
Epoch 23/50
12/12 - 1s - loss: 0.5853 - acc: 0.7917 - val_loss: 0.7640 - val_acc: 0.6111
Epoch 24/50
12/12 - 1s - loss: 0.3524 - acc: 0.8732 - val_loss: 0.6365 - val_acc: 0.7778
Epoch 25/50
12/12 - 1s - loss: 0.3922 - acc: 0.8750 - val_loss: 0.4557 - val_acc: 0.8333
Epoch 26/50
12/12 - 1s - loss: 0.4058 - acc: 0.8873 - val_loss: 0.9184 - val_acc: 0.6111
Epoch 27/50
12/12 - 1s - loss: 0.2918 - acc: 0.8873 - val_loss: 0.5392 - val_acc: 0.8889
Epoch 28/50
12/12 - 1s - loss: 0.2852 - acc: 0.9155 - val_loss: 0.8592 - val_acc: 0.7222
Epoch 29/50
12/12 - 1s - loss: 0.1907 - acc: 0.9155 - val_loss: 0.6888 - val_acc: 0.7778
Epoch 30/50
12/12 - 1s - loss: 0.2106 - acc: 0.9296 - val_loss: 0.7497 - val_acc: 0.8333
Epoch 31/50
12/12 - 1s - loss: 0.1489 - acc: 0.9444 - val_loss: 0.7338 - val_acc: 0.7222
Epoch 32/50
12/12 - 1s - loss: 0.1640 - acc: 0.9577 - val_loss: 0.7759 - val_acc: 0.8333
Epoch 33/50
12/12 - 1s - loss: 0.0594 - acc: 0.9859 - val_loss: 0.9707 - val_acc: 0.7778
Epoch 34/50
12/12 - 1s - loss: 0.1665 - acc: 0.9577 - val_loss: 0.8778 - val_acc: 0.8333
Epoch 35/50
12/12 - 1s - loss: 0.1040 - acc: 0.9722 - val_loss: 0.8476 - val_acc: 0.8333

=-====================== best model 
[True, False, False, False, True, True, True, False, True, True, True, True, True, True, True, True, False, False, True, True, True, True, True, True]
18
24
0.75
======================= end save
19
24
0.7916666666666666

=============train_datagen & val_datagen batch_size=6
12/12 - 3s - loss: 1.5069 - acc: 0.2113 - val_loss: 1.3843 - val_acc: 0.2778
Epoch 2/50
12/12 - 1s - loss: 1.4093 - acc: 0.2254 - val_loss: 1.3933 - val_acc: 0.3333
Epoch 3/50
12/12 - 1s - loss: 1.3555 - acc: 0.3472 - val_loss: 1.3699 - val_acc: 0.2222
Epoch 4/50
12/12 - 1s - loss: 1.3288 - acc: 0.2676 - val_loss: 1.3160 - val_acc: 0.3889
Epoch 5/50
12/12 - 1s - loss: 1.2926 - acc: 0.4085 - val_loss: 1.2605 - val_acc: 0.6111
Epoch 6/50
12/12 - 1s - loss: 1.1319 - acc: 0.5915 - val_loss: 1.1348 - val_acc: 0.5000
Epoch 7/50
12/12 - 1s - loss: 1.0380 - acc: 0.5972 - val_loss: 0.8520 - val_acc: 0.6667
Epoch 8/50
12/12 - 1s - loss: 0.8424 - acc: 0.6197 - val_loss: 1.1161 - val_acc: 0.3889
Epoch 9/50
12/12 - 1s - loss: 0.8065 - acc: 0.6250 - val_loss: 0.7445 - val_acc: 0.7778
Epoch 10/50
12/12 - 1s - loss: 0.6401 - acc: 0.7183 - val_loss: 0.5546 - val_acc: 0.9444
Epoch 11/50
12/12 - 1s - loss: 0.6773 - acc: 0.7042 - val_loss: 1.1341 - val_acc: 0.5000
Epoch 12/50
12/12 - 1s - loss: 0.7293 - acc: 0.6338 - val_loss: 1.0154 - val_acc: 0.5556
Epoch 13/50
12/12 - 1s - loss: 0.7866 - acc: 0.6761 - val_loss: 0.8060 - val_acc: 0.7778
Epoch 14/50
12/12 - 1s - loss: 0.4485 - acc: 0.8333 - val_loss: 0.8920 - val_acc: 0.7222
Epoch 15/50
12/12 - 1s - loss: 0.3676 - acc: 0.8169 - val_loss: 0.5344 - val_acc: 0.7778
Epoch 16/50
12/12 - 1s - loss: 0.4782 - acc: 0.8592 - val_loss: 0.7138 - val_acc: 0.7222
Epoch 17/50
12/12 - 1s - loss: 0.3980 - acc: 0.8592 - val_loss: 0.6684 - val_acc: 0.7778
Epoch 18/50
12/12 - 1s - loss: 0.2433 - acc: 0.9296 - val_loss: 0.6507 - val_acc: 0.7778
Epoch 19/50
12/12 - 1s - loss: 0.1552 - acc: 0.9722 - val_loss: 0.5813 - val_acc: 0.7778
Epoch 20/50
12/12 - 1s - loss: 0.2698 - acc: 0.8873 - val_loss: 0.6552 - val_acc: 0.7778
Epoch 21/50
12/12 - 1s - loss: 0.1110 - acc: 0.9577 - val_loss: 0.6749 - val_acc: 0.7778
Epoch 22/50
12/12 - 1s - loss: 0.1377 - acc: 0.9583 - val_loss: 0.7574 - val_acc: 0.7222
Epoch 23/50
12/12 - 1s - loss: 0.0992 - acc: 0.9577 - val_loss: 0.5383 - val_acc: 0.8333
Epoch 24/50
12/12 - 1s - loss: 0.1378 - acc: 0.9296 - val_loss: 0.6087 - val_acc: 0.7778
Epoch 25/50
12/12 - 1s - loss: 0.0415 - acc: 1.0000 - val_loss: 0.5924 - val_acc: 0.8333

=-====================== best model 

======================= end save
19
24
0.7916666666666666

=============train_datagen & val_datagen batch_size=6

Epoch 1/50

  "Palette images with Transparency expressed in bytes should be "
12/12 - 3s - loss: 1.5876 - acc: 0.2394 - val_loss: 1.3829 - val_acc: 0.3333
Epoch 2/50
12/12 - 1s - loss: 1.4057 - acc: 0.2113 - val_loss: 1.3847 - val_acc: 0.3333
Epoch 3/50
12/12 - 1s - loss: 1.3950 - acc: 0.1944 - val_loss: 1.3838 - val_acc: 0.2778
Epoch 4/50
12/12 - 1s - loss: 1.3768 - acc: 0.2535 - val_loss: 1.3699 - val_acc: 0.2778
Epoch 5/50
12/12 - 1s - loss: 1.3094 - acc: 0.3472 - val_loss: 1.3312 - val_acc: 0.5556
Epoch 6/50
12/12 - 1s - loss: 1.3109 - acc: 0.4225 - val_loss: 1.1574 - val_acc: 0.5556
Epoch 7/50
12/12 - 1s - loss: 1.1116 - acc: 0.4930 - val_loss: 0.9733 - val_acc: 0.6111
Epoch 8/50
12/12 - 1s - loss: 1.1613 - acc: 0.5634 - val_loss: 1.1057 - val_acc: 0.5000
Epoch 9/50
12/12 - 1s - loss: 0.8463 - acc: 0.6620 - val_loss: 0.5694 - val_acc: 0.7778
Epoch 10/50
12/12 - 1s - loss: 0.5809 - acc: 0.7606 - val_loss: 0.6953 - val_acc: 0.6111
Epoch 11/50
12/12 - 1s - loss: 0.6425 - acc: 0.7324 - val_loss: 0.4259 - val_acc: 0.8333
Epoch 12/50
12/12 - 1s - loss: 0.6339 - acc: 0.7083 - val_loss: 0.8621 - val_acc: 0.6111
Epoch 13/50
12/12 - 1s - loss: 0.5373 - acc: 0.8169 - val_loss: 0.4460 - val_acc: 0.7778
Epoch 14/50
12/12 - 1s - loss: 0.5271 - acc: 0.8310 - val_loss: 0.6876 - val_acc: 0.6111
Epoch 15/50
12/12 - 1s - loss: 0.3585 - acc: 0.8472 - val_loss: 0.3363 - val_acc: 0.7778
Epoch 16/50
12/12 - 1s - loss: 0.2936 - acc: 0.9014 - val_loss: 0.4387 - val_acc: 0.8889
Epoch 17/50
12/12 - 1s - loss: 0.4390 - acc: 0.8592 - val_loss: 0.4149 - val_acc: 0.8889
Epoch 18/50
12/12 - 1s - loss: 0.2310 - acc: 0.9028 - val_loss: 0.5842 - val_acc: 0.8333
Epoch 19/50
12/12 - 1s - loss: 0.2983 - acc: 0.8873 - val_loss: 0.3888 - val_acc: 0.8333
Epoch 20/50
12/12 - 1s - loss: 0.3485 - acc: 0.9155 - val_loss: 0.7428 - val_acc: 0.7778
Epoch 21/50
12/12 - 1s - loss: 0.2237 - acc: 0.9028 - val_loss: 0.5091 - val_acc: 0.8333
Epoch 22/50
12/12 - 1s - loss: 0.1417 - acc: 0.9429 - val_loss: 0.2866 - val_acc: 0.9444
Epoch 23/50
12/12 - 1s - loss: 0.1124 - acc: 0.9722 - val_loss: 0.1975 - val_acc: 0.8889
Epoch 24/50
12/12 - 1s - loss: 0.0578 - acc: 0.9718 - val_loss: 0.1934 - val_acc: 0.8889
Epoch 25/50
12/12 - 1s - loss: 0.0319 - acc: 1.0000 - val_loss: 0.1842 - val_acc: 0.9444
Epoch 26/50
12/12 - 1s - loss: 0.0272 - acc: 1.0000 - val_loss: 0.2437 - val_acc: 0.9444
Epoch 27/50
12/12 - 1s - loss: 0.0182 - acc: 1.0000 - val_loss: 0.1499 - val_acc: 0.9444
Epoch 28/50
12/12 - 1s - loss: 0.0042 - acc: 1.0000 - val_loss: 0.0443 - val_acc: 1.0000
Epoch 29/50
12/12 - 1s - loss: 0.0140 - acc: 1.0000 - val_loss: 0.1630 - val_acc: 0.9444
Epoch 30/50
12/12 - 1s - loss: 0.2959 - acc: 0.9296 - val_loss: 0.4106 - val_acc: 0.9444
Epoch 31/50
12/12 - 1s - loss: 0.1117 - acc: 0.9583 - val_loss: 0.5020 - val_acc: 0.7778
Epoch 32/50
12/12 - 1s - loss: 0.1862 - acc: 0.9155 - val_loss: 0.2083 - val_acc: 0.8333
Epoch 33/50
12/12 - 1s - loss: 0.2660 - acc: 0.9437 - val_loss: 0.2789 - val_acc: 0.8889
Epoch 34/50
12/12 - 1s - loss: 0.1294 - acc: 0.9577 - val_loss: 0.1664 - val_acc: 0.8889
Epoch 35/50
12/12 - 1s - loss: 0.0249 - acc: 1.0000 - val_loss: 0.0311 - val_acc: 1.0000
Epoch 36/50
12/12 - 1s - loss: 0.0093 - acc: 1.0000 - val_loss: 0.0409 - val_acc: 1.0000
Epoch 37/50
12/12 - 1s - loss: 0.0324 - acc: 1.0000 - val_loss: 0.1031 - val_acc: 0.9444
Epoch 38/50
12/12 - 1s - loss: 0.0434 - acc: 0.9722 - val_loss: 0.0432 - val_acc: 1.0000
Epoch 39/50
12/12 - 1s - loss: 0.0292 - acc: 0.9857 - val_loss: 0.0669 - val_acc: 1.0000
Epoch 40/50
12/12 - 1s - loss: 0.0844 - acc: 0.9861 - val_loss: 0.6314 - val_acc: 0.7222
Epoch 41/50
12/12 - 1s - loss: 0.0740 - acc: 0.9577 - val_loss: 0.4387 - val_acc: 0.9444
Epoch 42/50
12/12 - 1s - loss: 0.0439 - acc: 0.9859 - val_loss: 0.3860 - val_acc: 0.8889
Epoch 43/50
12/12 - 1s - loss: 0.0084 - acc: 1.0000 - val_loss: 0.2996 - val_acc: 0.8889
Epoch 44/50
12/12 - 1s - loss: 0.0189 - acc: 1.0000 - val_loss: 0.1878 - val_acc: 0.9444
Epoch 45/50
12/12 - 1s - loss: 0.0168 - acc: 0.9861 - val_loss: 0.1696 - val_acc: 0.9444

=-====================== best model 
22
24
0.9166666666666666
======================= end save
21
24
0.875

=============train_datagen & val_datagen batch_size=4

Epoch 2/50
12/12 - 1s - loss: 1.3928 - acc: 0.3617 - val_loss: 1.3793 - val_acc: 0.3333
Epoch 3/50
12/12 - 1s - loss: 1.3901 - acc: 0.1702 - val_loss: 1.3784 - val_acc: 0.3333
Epoch 4/50
12/12 - 1s - loss: 1.3753 - acc: 0.2917 - val_loss: 1.3698 - val_acc: 0.3333
Epoch 5/50
12/12 - 1s - loss: 1.3578 - acc: 0.3191 - val_loss: 1.3500 - val_acc: 0.3333
Epoch 6/50
12/12 - 1s - loss: 1.4046 - acc: 0.1875 - val_loss: 1.3687 - val_acc: 0.3333
Epoch 7/50
12/12 - 1s - loss: 1.3488 - acc: 0.3958 - val_loss: 1.3558 - val_acc: 0.5833
Epoch 8/50
12/12 - 1s - loss: 1.3282 - acc: 0.2553 - val_loss: 1.3121 - val_acc: 0.5000
Epoch 9/50
12/12 - 1s - loss: 1.2393 - acc: 0.5000 - val_loss: 1.1969 - val_acc: 0.4167
Epoch 10/50
12/12 - 1s - loss: 1.1807 - acc: 0.4681 - val_loss: 1.2825 - val_acc: 0.3333
Epoch 11/50
12/12 - 1s - loss: 1.1404 - acc: 0.4583 - val_loss: 1.1595 - val_acc: 0.4167
Epoch 12/50
12/12 - 1s - loss: 0.7377 - acc: 0.7660 - val_loss: 1.1961 - val_acc: 0.4167
Epoch 13/50
12/12 - 1s - loss: 0.9723 - acc: 0.6170 - val_loss: 1.1051 - val_acc: 0.5000
Epoch 14/50
12/12 - 1s - loss: 0.8133 - acc: 0.6458 - val_loss: 1.1023 - val_acc: 0.6667
Epoch 15/50
12/12 - 1s - loss: 0.6107 - acc: 0.7872 - val_loss: 1.0351 - val_acc: 0.6667
Epoch 16/50
12/12 - 1s - loss: 0.7098 - acc: 0.7500 - val_loss: 0.9534 - val_acc: 0.6667
Epoch 17/50
12/12 - 1s - loss: 0.4934 - acc: 0.8085 - val_loss: 0.7624 - val_acc: 0.8333
Epoch 18/50
12/12 - 1s - loss: 0.3775 - acc: 0.8750 - val_loss: 1.0094 - val_acc: 0.5833
Epoch 19/50
12/12 - 1s - loss: 0.3040 - acc: 0.8958 - val_loss: 0.7662 - val_acc: 0.5833
Epoch 20/50
12/12 - 1s - loss: 0.4360 - acc: 0.8085 - val_loss: 0.9694 - val_acc: 0.6667
Epoch 21/50
12/12 - 1s - loss: 0.5942 - acc: 0.7917 - val_loss: 1.2980 - val_acc: 0.4167
Epoch 22/50
12/12 - 1s - loss: 0.3473 - acc: 0.8511 - val_loss: 1.1692 - val_acc: 0.6667
Epoch 23/50
12/12 - 1s - loss: 0.4162 - acc: 0.8958 - val_loss: 0.9021 - val_acc: 0.6667
Epoch 24/50
12/12 - 1s - loss: 0.2644 - acc: 0.9362 - val_loss: 0.8402 - val_acc: 0.7500
Epoch 25/50
12/12 - 1s - loss: 0.2067 - acc: 0.8723 - val_loss: 1.0619 - val_acc: 0.5833
Epoch 26/50
12/12 - 1s - loss: 0.1780 - acc: 0.9375 - val_loss: 1.1045 - val_acc: 0.7500
Epoch 27/50
12/12 - 1s - loss: 0.0893 - acc: 0.9787 - val_loss: 1.0078 - val_acc: 0.7500


=-====================== best model save(17/50)

[True, True, True, False, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
21
24
0.875


======================= end save

19
24
0.7916666666666666


=============train_datagen & val_datagen batch_size=2

12/12 - 0s - loss: 1.3823 - acc: 0.4348 - val_loss: 1.3997 - val_acc: 0.1667
Epoch 3/50
12/12 - 0s - loss: 1.3769 - acc: 0.3750 - val_loss: 1.3975 - val_acc: 0.1667
Epoch 4/50
12/12 - 0s - loss: 1.3675 - acc: 0.3333 - val_loss: 1.4078 - val_acc: 0.1667
Epoch 5/50
12/12 - 0s - loss: 1.3807 - acc: 0.4167 - val_loss: 1.4257 - val_acc: 0.1667
Epoch 6/50
12/12 - 0s - loss: 1.4356 - acc: 0.2500 - val_loss: 1.4040 - val_acc: 0.1667
Epoch 7/50
12/12 - 0s - loss: 1.4038 - acc: 0.3043 - val_loss: 1.3947 - val_acc: 0.1667
Epoch 8/50
12/12 - 0s - loss: 1.4092 - acc: 0.2083 - val_loss: 1.3911 - val_acc: 0.1667
Epoch 9/50
12/12 - 0s - loss: 1.3754 - acc: 0.4348 - val_loss: 1.3929 - val_acc: 0.1667
Epoch 10/50
12/12 - 0s - loss: 1.3995 - acc: 0.1667 - val_loss: 1.3959 - val_acc: 0.1667
Epoch 11/50
12/12 - 0s - loss: 1.3888 - acc: 0.3333 - val_loss: 1.3947 - val_acc: 0.1667
Epoch 12/50
12/12 - 0s - loss: 1.3868 - acc: 0.3333 - val_loss: 1.3829 - val_acc: 0.1667
Epoch 13/50
12/12 - 0s - loss: 1.3932 - acc: 0.1667 - val_loss: 1.3865 - val_acc: 0.3333
Epoch 14/50
12/12 - 0s - loss: 1.3873 - acc: 0.2174 - val_loss: 1.3809 - val_acc: 0.1667
Epoch 15/50
12/12 - 0s - loss: 1.3967 - acc: 0.1667 - val_loss: 1.3839 - val_acc: 0.1667
Epoch 16/50
12/12 - 0s - loss: 1.3866 - acc: 0.2917 - val_loss: 1.3856 - val_acc: 0.1667
Epoch 17/50
12/12 - 0s - loss: 1.3895 - acc: 0.2917 - val_loss: 1.3885 - val_acc: 0.1667
Epoch 18/50
12/12 - 0s - loss: 1.3070 - acc: 0.5000 - val_loss: 1.3898 - val_acc: 0.1667
Epoch 19/50
12/12 - 0s - loss: 1.3464 - acc: 0.4348 - val_loss: 1.3557 - val_acc: 0.1667
Epoch 20/50
12/12 - 0s - loss: 1.5903 - acc: 0.2917 - val_loss: 1.3877 - val_acc: 0.1667
Epoch 21/50
12/12 - 0s - loss: 1.3882 - acc: 0.3478 - val_loss: 1.3939 - val_acc: 0.1667
Epoch 22/50
12/12 - 0s - loss: 1.3913 - acc: 0.2917 - val_loss: 1.3986 - val_acc: 0.1667
Epoch 23/50
12/12 - 0s - loss: 1.3882 - acc: 0.1667 - val_loss: 1.4000 - val_acc: 0.1667
Epoch 24/50
12/12 - 0s - loss: 1.3882 - acc: 0.2083 - val_loss: 1.3968 - val_acc: 0.1667
Epoch 25/50
12/12 - 0s - loss: 1.3870 - acc: 0.2609 - val_loss: 1.3828 - val_acc: 0.1667
Epoch 26/50
12/12 - 0s - loss: 1.3983 - acc: 0.1250 - val_loss: 1.3809 - val_acc: 0.1667
Epoch 27/50
12/12 - 0s - loss: 1.3973 - acc: 0.1667 - val_loss: 1.3941 - val_acc: 0.1667
Epoch 28/50
12/12 - 0s - loss: 1.3925 - acc: 0.2083 - val_loss: 1.3964 - val_acc: 0.1667
Epoch 29/50
12/12 - 0s - loss: 1.3991 - acc: 0.2083 - val_loss: 1.3892 - val_acc: 0.1667

=-====================== best model 

[True, True, True, False, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
6
24
0.25
======================= end save
6
24
0.25

=============train_datagen & val_datagen batch_size=2

Epoch 1/50

  "Palette images with Transparency expressed in bytes should be "
12/12 - 2s - loss: 1.6791 - acc: 0.3333 - val_loss: 1.3767 - val_acc: 0.1667
Epoch 2/50
12/12 - 0s - loss: 1.4143 - acc: 0.3043 - val_loss: 1.4143 - val_acc: 0.1667
Epoch 3/50
12/12 - 0s - loss: 1.4167 - acc: 0.2083 - val_loss: 1.4089 - val_acc: 0.1667
Epoch 4/50
12/12 - 0s - loss: 1.4170 - acc: 0.2917 - val_loss: 1.3932 - val_acc: 0.1667
Epoch 5/50
12/12 - 0s - loss: 1.3849 - acc: 0.3478 - val_loss: 1.3893 - val_acc: 0.1667
Epoch 6/50
12/12 - 0s - loss: 1.3777 - acc: 0.2917 - val_loss: 1.3910 - val_acc: 0.1667
Epoch 7/50
12/12 - 0s - loss: 1.3930 - acc: 0.2917 - val_loss: 1.3835 - val_acc: 0.1667
Epoch 8/50
12/12 - 0s - loss: 1.3928 - acc: 0.2083 - val_loss: 1.3859 - val_acc: 0.1667
Epoch 9/50
12/12 - 0s - loss: 1.3880 - acc: 0.2917 - val_loss: 1.3813 - val_acc: 0.1667
Epoch 10/50
12/12 - 0s - loss: 1.3379 - acc: 0.2917 - val_loss: 1.3570 - val_acc: 0.1667
Epoch 11/50
12/12 - 0s - loss: 1.4118 - acc: 0.3913 - val_loss: 1.3743 - val_acc: 0.3333
Epoch 12/50
12/12 - 0s - loss: 1.3666 - acc: 0.3333 - val_loss: 1.3321 - val_acc: 0.6667
Epoch 13/50
12/12 - 0s - loss: 1.2393 - acc: 0.5000 - val_loss: 1.1981 - val_acc: 0.1667
Epoch 14/50
12/12 - 0s - loss: 1.4492 - acc: 0.5417 - val_loss: 1.4039 - val_acc: 0.1667
Epoch 15/50
12/12 - 0s - loss: 1.3842 - acc: 0.3913 - val_loss: 1.3990 - val_acc: 0.1667
Epoch 16/50
12/12 - 0s - loss: 1.2615 - acc: 0.5833 - val_loss: 1.3070 - val_acc: 0.3333
Epoch 17/50
12/12 - 0s - loss: 1.1195 - acc: 0.5000 - val_loss: 1.1786 - val_acc: 0.6667
Epoch 18/50
12/12 - 0s - loss: 1.1330 - acc: 0.4583 - val_loss: 1.1987 - val_acc: 0.8333
Epoch 19/50
12/12 - 0s - loss: 1.0055 - acc: 0.4783 - val_loss: 1.2311 - val_acc: 0.6667
Epoch 20/50
12/12 - 0s - loss: 1.0824 - acc: 0.7500 - val_loss: 1.0195 - val_acc: 0.6667
Epoch 21/50
12/12 - 0s - loss: 0.8234 - acc: 0.6522 - val_loss: 0.8360 - val_acc: 1.0000
Epoch 22/50
12/12 - 0s - loss: 0.8560 - acc: 0.6667 - val_loss: 0.9709 - val_acc: 0.5000
Epoch 23/50
12/12 - 0s - loss: 0.9123 - acc: 0.6250 - val_loss: 1.0914 - val_acc: 0.5000
Epoch 24/50
12/12 - 0s - loss: 0.7036 - acc: 0.6250 - val_loss: 0.9543 - val_acc: 0.5000
Epoch 25/50
12/12 - 0s - loss: 0.8375 - acc: 0.6522 - val_loss: 0.9577 - val_acc: 0.3333
Epoch 26/50
12/12 - 0s - loss: 0.7323 - acc: 0.6667 - val_loss: 1.2937 - val_acc: 0.5000
Epoch 27/50
12/12 - 0s - loss: 0.9130 - acc: 0.7500 - val_loss: 0.9084 - val_acc: 0.6667
Epoch 28/50
12/12 - 0s - loss: 1.1023 - acc: 0.5000 - val_loss: 1.1919 - val_acc: 0.6667
Epoch 29/50
12/12 - 0s - loss: 0.8694 - acc: 0.6250 - val_loss: 0.8986 - val_acc: 0.6667
Epoch 30/50
12/12 - 0s - loss: 0.5457 - acc: 0.7826 - val_loss: 0.7224 - val_acc: 0.8333
Epoch 31/50
12/12 - 0s - loss: 0.6862 - acc: 0.7500 - val_loss: 1.4066 - val_acc: 0.3333
Epoch 32/50
12/12 - 0s - loss: 0.5417 - acc: 0.7917 - val_loss: 1.6421 - val_acc: 0.1667
Epoch 33/50
12/12 - 0s - loss: 0.6835 - acc: 0.6667 - val_loss: 0.9678 - val_acc: 0.5000
Epoch 34/50
12/12 - 0s - loss: 0.6058 - acc: 0.7083 - val_loss: 0.8739 - val_acc: 0.6667
Epoch 35/50
12/12 - 0s - loss: 0.5164 - acc: 0.7500 - val_loss: 0.9773 - val_acc: 0.5000
Epoch 36/50
12/12 - 0s - loss: 0.9135 - acc: 0.6957 - val_loss: 0.7827 - val_acc: 0.6667
Epoch 37/50
12/12 - 0s - loss: 0.6138 - acc: 0.7917 - val_loss: 0.9902 - val_acc: 0.3333
Epoch 38/50
12/12 - 0s - loss: 0.3973 - acc: 0.8696 - val_loss: 0.9536 - val_acc: 0.3333
Epoch 39/50
12/12 - 0s - loss: 0.4686 - acc: 0.7917 - val_loss: 0.7066 - val_acc: 0.8333
Epoch 40/50
12/12 - 0s - loss: 0.4339 - acc: 0.8333 - val_loss: 0.7634 - val_acc: 0.6667
Epoch 41/50
12/12 - 0s - loss: 0.4540 - acc: 0.7917 - val_loss: 0.8571 - val_acc: 0.8333
Epoch 42/50
12/12 - 0s - loss: 0.2571 - acc: 0.9167 - val_loss: 0.9269 - val_acc: 0.6667
Epoch 43/50
12/12 - 0s - loss: 0.6589 - acc: 0.7917 - val_loss: 0.9175 - val_acc: 0.5000
Epoch 44/50
12/12 - 0s - loss: 0.2481 - acc: 0.9130 - val_loss: 1.0063 - val_acc: 0.3333
Epoch 45/50
12/12 - 0s - loss: 0.3807 - acc: 0.8696 - val_loss: 0.8532 - val_acc: 0.8333
Epoch 46/50
12/12 - 0s - loss: 0.1264 - acc: 1.0000 - val_loss: 0.7736 - val_acc: 0.8333
Epoch 47/50
12/12 - 0s - loss: 0.5784 - acc: 0.7917 - val_loss: 0.7685 - val_acc: 0.8333
Epoch 48/50
12/12 - 0s - loss: 0.2773 - acc: 0.9167 - val_loss: 0.7550 - val_acc: 0.6667
Epoch 49/50
12/12 - 0s - loss: 0.1175 - acc: 0.9583 - val_loss: 0.7300 - val_acc: 0.8333

=-====================== best model 

[True, True, True, False, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
18
24
0.75
======================= end save
18
24
0.75

=============train_datagen & val_datagen batch_size=2

Epoch 1/50

  "Palette images with Transparency expressed in bytes should be "
12/12 - 2s - loss: 1.7255 - acc: 0.2500 - val_loss: 1.3907 - val_acc: 0.3333
Epoch 2/50
12/12 - 0s - loss: 1.4694 - acc: 0.1250 - val_loss: 1.3921 - val_acc: 0.3333
Epoch 3/50
12/12 - 0s - loss: 1.3893 - acc: 0.3043 - val_loss: 1.3805 - val_acc: 0.3333
Epoch 4/50
12/12 - 0s - loss: 1.3923 - acc: 0.2500 - val_loss: 1.3763 - val_acc: 0.3333
Epoch 5/50
12/12 - 0s - loss: 1.3959 - acc: 0.2500 - val_loss: 1.3770 - val_acc: 0.3333
Epoch 6/50
12/12 - 0s - loss: 1.3912 - acc: 0.2083 - val_loss: 1.3813 - val_acc: 0.3333
Epoch 7/50
12/12 - 0s - loss: 1.3856 - acc: 0.2500 - val_loss: 1.3852 - val_acc: 0.3333
Epoch 8/50
12/12 - 0s - loss: 1.3997 - acc: 0.2174 - val_loss: 1.3775 - val_acc: 0.3333
Epoch 9/50
12/12 - 0s - loss: 1.3811 - acc: 0.2500 - val_loss: 1.3796 - val_acc: 0.3333
Epoch 10/50
12/12 - 0s - loss: 1.3649 - acc: 0.1667 - val_loss: 1.3493 - val_acc: 0.6667
Epoch 11/50
12/12 - 0s - loss: 1.4336 - acc: 0.2917 - val_loss: 1.3710 - val_acc: 0.3333
Epoch 12/50
12/12 - 0s - loss: 1.3992 - acc: 0.1304 - val_loss: 1.3943 - val_acc: 0.3333
Epoch 13/50
12/12 - 0s - loss: 1.3778 - acc: 0.3043 - val_loss: 1.3939 - val_acc: 0.3333
Epoch 14/50
12/12 - 0s - loss: 1.3803 - acc: 0.4583 - val_loss: 1.3891 - val_acc: 0.3333
Epoch 15/50
12/12 - 0s - loss: 1.3327 - acc: 0.2500 - val_loss: 1.4245 - val_acc: 0.5000
Epoch 16/50
12/12 - 0s - loss: 1.5164 - acc: 0.1667 - val_loss: 1.3931 - val_acc: 0.3333
Epoch 17/50
12/12 - 0s - loss: 1.4048 - acc: 0.2174 - val_loss: 1.3877 - val_acc: 0.3333
Epoch 18/50
12/12 - 0s - loss: 1.3925 - acc: 0.2917 - val_loss: 1.3897 - val_acc: 0.3333
Epoch 19/50
12/12 - 0s - loss: 1.3962 - acc: 0.1667 - val_loss: 1.3906 - val_acc: 0.3333
Epoch 20/50
12/12 - 0s - loss: 1.3874 - acc: 0.2917 - val_loss: 1.3932 - val_acc: 0.3333

=-====================== best model 

[True, True, True, False, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True]
11
24
0.4583333333333333

======================= end save
0.25