adjust super params and mk dir for restoring models

add train_inceptionV3_rawimg.py using raw img as input; adjust float to unit8 for train_inceptionV3.py; add predict for inceptionV3; update README.md

add train_reload_inceptionV3

add inceptionV3 pretrain model

remove train_mode placeholder because already using keep_prob

pass keep_prob to vgg19_trainable.py

add vgg19_trainable

fix y_ name error; adjust RELOAD_BASE to 10000 and keep_prob to 0.9

adjust GradientDescentOptimizer to AdamOptimizer in alexnet to make fitting

add alexnet with lr decay

tidy up predict.py; update README.md for train_vgg.py & train_reload.py

name y_ in train_cnn & train_vgg16; add train_reload.py for continue trainging based on RELOAD_BASE node

add vgg19.py for vgg19 net

add trainin_vgg16 based on vgg16 model; modify x from [WIDTH*HEIGHT*CHANNEL] to [WIDTH, HEIGHT, CHANNEL]

add model reload for predict; adjust conv weight; save model by SAVE_MODEL_INTERVAL