def __init__(self, step):
with tf.variable_scope('Input_holders'):
self.inpholder = tf.placeholder(tf.float32,
[None, step, 256, 256, 3])
self.step = step
self.targetholder = tf.placeholder(tf.float32,
[None, step, 256, 256, 3])
with tf.variable_scope('Head'):
inp_split = tf.unstack(self.inpholder, axis=1)
features = []
for i in range(len(inp_split)):
features.append(N.conv_layers(inp_split[i], i != 0))
features = tf.stack(features, axis=1)
lstm_out = M.SimpleLSTM(4 * 4 * 128).apply(features)
with tf.variable_scope('Tail'):
feat_split = tf.unstack(lstm_out, axis=1)
# I try the last frame for now
feat = feat_split[-1]
A, C = N.deconv_layers(feat)
self.recon = A * C + (1. - A) * self.inpholder[:, -1]
self.A = A
self.C = C
self.build_loss()
self.saver = tf.train.Saver()
self.sess = tf.Session()
M.loadSess('./model/', self.sess, init=True)
def main():
holders, losses, upops, trains, x_fake, lb_ls = build_graph()
saver = tf.train.Saver()
with tf.Session() as sess:
M.loadSess('./model/', sess, init=True)
data = read_data()
for iteration in range(10000):
train_batch = random.sample(data, BSIZE)
img_batch = [i[0] for i in train_batch]
gnd_batch = [i[1] for i in train_batch]
feeddict = {holders[0]: img_batch, holders[1]: gnd_batch}
g_loss, d_loss, _, _, _, _, lb_g, lb_d = sess.run(
losses + upops + trains + lb_ls, feed_dict=feeddict)
print('Iter:', iteration, '\tLoss_g:', g_loss, '\tLb_g:', lb_g,
'\tLoss_d:', d_loss)
while lb_g > lb_d + 0.7:
train_batch = random.sample(data, BSIZE)
img_batch = [i[0] for i in train_batch]
gnd_batch = [i[1] for i in train_batch]
feeddict = {holders[0]: img_batch, holders[1]: gnd_batch}
g_loss, d_loss, _, _, lb_g, lb_d = sess.run(
losses + [upops[0], trains[0]] + lb_ls, feed_dict=feeddict)
print('Iter:', iteration, '\tLoss_g:', g_loss, '\tLb_g:', lb_g,
'\tLoss_d:', d_loss)
if iteration % 100 == 0 and iteration > 0:
saver.save(sess, './model/' + str(iteration) + '.ckpt')
if iteration % 20 == 0:
gen = sess.run(x_fake, feed_dict=feeddict)
show_sample(gen, img_batch, gnd_batch, iteration)
def __init__(self):
with tf.variable_scope('holders'):
self.inp_holder = tf.placeholder(tf.float32, [None, 28, 28, 1])
self.lab_holder = tf.placeholder(tf.float32, [None, 10])
with tf.variable_scope('mainMod'):
mod = M.Model(self.inp_holder)
mod.convLayer(7, 64, stride=2, activation=M.PARAM_RELU)
mod.convLayer(5, 128, stride=2, activation=M.PARAM_RELU)
mod.capsulization(dim=16, caps=8)
mod.caps_conv(3, 8, 16, activation=None, usebias=False)
mod.caps_flatten()
mod.squash()
mod.capsLayer(10, 8, 3, BSIZE=128)
mod.squash()
feat = mod.capsDown()
with tf.variable_scope('loss'):
length = tf.norm(feat, axis=2)
self.length = length
loss = self.lab_holder * tf.square(tf.maximum(
0., 0.9 - length)) + 0.5 * (1 - self.lab_holder) * tf.square(
tf.maximum(0., length - 0.1))
self.loss = tf.reduce_mean(tf.reduce_sum(loss, 1))
self.accuracy = M.accuracy(length, tf.argmax(self.lab_holder, 1))
with tf.variable_scope('opti'):
self.train_op = tf.train.AdamOptimizer(0.001).minimize(self.loss)
self.sess = tf.Session()
M.loadSess(self.sess, './model/', init=True)
def training():
with tf.Session() as sess:
merged = tf.summary.merge_all()
print('creating log file...')
writer = tf.summary.FileWriter('./logs/', sess.graph)
saver = tf.train.Saver()
M.loadSess(modelpath, sess=sess)
imgs = list(getImgs())
print('start training...')
for i in range(MAXITER):
a = np.random.uniform(size=[BSIZE, ZDIM], low=-1.0, high=1.0)
# for _ in range(3):
# sess.run(trainG,feed_dict={z:a})
_, mg, lsd, lsg = sess.run(
[trainAll, merged, lossD, lossG, UPDD, UPDG],
feed_dict={
z: a,
imgholder: random.sample(imgs, BSIZE)
})
if (i) % 1 == 0:
writer.add_summary(mg, i)
print('iter:', i)
print('lsd:', lsd)
print('lsg:', lsg)
if (i + 1) % 100 == 0:
getGeneratedImg(sess, i + 1)
if (i + 1) % 1000 == 0:
saver.save(
sess, modelpath + 'Model_epoc' + str(i + 1) + 'Time' +
datetime.now().strftime('%Y%m%d%H%M%S') + '.ckpt')
def __init__(self, class_num):
inp_holder = tf.placeholder(tf.float32, [None, 460, 460, 3])
lab_holder = tf.placeholder(tf.int32, [None, 460, 460])
mask_holder = tf.placeholder(tf.float32, [None, 460, 460])
mask = tf.expand_dims(mask_holder, -1)
c_ = tf.concat([inp_holder, mask], -1)
merged_layer = self.merging_layer(c_)
self.net_body = seg_main_body(merged_layer)
seg_layer = self.segmentation_layer(self.net_body.feature_layer, 12,
class_num)
self.build_loss(seg_layer, lab_holder)
self.saver = tf.train.Saver()
self.sess = tf.Session()
M.loadSess('./model/',
self.sess,
init=True,
var_list=self.net_body.var)
self.inp_holder = inp_holder
self.lab_holder = lab_holder
self.seg_layer = seg_layer
self.mask_holder = mask_holder
def __init__(self, class_num, is_training=True, mod_dir='./model/'):
self.mod_dir = mod_dir
with tf.variable_scope('Input'):
self.img_holder = tf.placeholder(tf.float32, [None, 128, 128, 3])
self.lab_holder = tf.placeholder(tf.float32, [None, class_num])
with tf.variable_scope('Res_101_cy'):
mod = M.Model(self.img_holder)
mod.set_bn_training(is_training)
# 64x64
mod.convLayer(7,
64,
stride=2,
activation=M.PARAM_LRELU,
batch_norm=True)
mod.res_block(256, stride=1, activation=M.PARAM_LRELU)
mod.res_block(256, stride=1, activation=M.PARAM_LRELU)
mod.res_block(256, stride=1, activation=M.PARAM_LRELU)
# 32x32
mod.res_block(512, stride=2, activation=M.PARAM_LRELU)
mod.res_block(512, stride=1, activation=M.PARAM_LRELU)
mod.res_block(512, stride=1, activation=M.PARAM_LRELU)
mod.res_block(512, stride=1, activation=M.PARAM_LRELU)
# 16x16
for i in range(14):
mod.res_block(1024, stride=2, activation=M.PARAM_LRELU)
# 8x8
mod.res_block(2048, stride=2, activation=M.PARAM_LRELU)
mod.res_block(2048, stride=1, activation=M.PARAM_LRELU)
mod.res_block(2048, stride=1, activation=M.PARAM_LRELU)
mod.avgpoolLayer(8)
mod.flatten()
#mod.fcLayer(256,nobias=True)
self.feat = mod.get_current_layer()
with tf.variable_scope('Classification'):
logit_layer, eval_layer = M.enforcedClassifier(self.feat,
self.lab_holder,
dropout=1,
multi=None,
L2norm=False)
self.accuracy = M.accuracy(eval_layer,
tf.argmax(self.lab_holder, -1))
if is_training:
print('Building optimizer...')
self.loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(
logits=logit_layer, labels=self.lab_holder))
with tf.control_dependencies(M.get_update_ops()):
self.train_op = tf.train.AdamOptimizer(0.0001).minimize(
self.loss)
self.sess = tf.Session()
self.saver = tf.train.Saver()
M.loadSess(mod_dir, self.sess, init=True)
def __init__(self, age_size, id_num, model_path='./aim_model/'):
self.model_path = model_path
self.inp_holder = tf.placeholder(tf.float32, [None, 128, 128, 3])
# self.real_holder = tf.placeholder(tf.float32,[None,128,128,3])
self.uni_holder = tf.placeholder(tf.float32, [None, 2, 2, 512])
self.age_holder = tf.placeholder(tf.float32, [None, age_size])
self.target_holder = tf.placeholder(tf.float32, [None, 128, 128, 3])
self.id_holder = tf.placeholder(tf.float32, [None, id_num])
# get_feature
self.feat = N.feat_encoder(self.inp_holder)
# retrieve tensor for adv1 and ip
adv1, ip = N.discriminator_f(self.feat, id_num)
adv1_uni, _ = N.discriminator_f(self.uni_holder, id_num)
# get attention A and C
age_expanded = self.expand(self.age_holder, self.feat)
aged_feature = tf.concat([age_expanded, self.feat], -1)
self.A, self.C = N.generator_att(aged_feature)
# construct synthesized image
self.generated = self.A * self.C + (1. - self.A) * self.inp_holder
# retrieve tensor for adv2 and ae
adv2, age_pred = N.discriminator(self.generated, age_size)
adv2_real, age_pred_real = N.discriminator(self.target_holder,
age_size)
# retrieve tensor for ai1 and ai2
ai1 = N.age_classify_r(self.feat, age_size)
ai2 = N.age_classify(self.feat, age_size)
# call loss builder functions
print('Building losses...')
self.build_loss_mc()
self.build_loss_adv1(adv1, adv1_uni)
self.build_loss_ip(ip)
self.build_loss_adv2(adv2, adv2_real)
self.build_loss_ae(age_pred, age_pred_real)
self.build_loss_ai1(ai1)
self.build_loss_ai2(ai2, age_size)
self.build_loss_A()
self.update_ops()
self.accuracy = M.accuracy(ip, tf.argmax(self.id_holder, -1))
self.sess = tf.Session()
M.loadSess(model_path, self.sess, init=True)
self.saver = tf.train.Saver()
def __init__(self):
inp_holder = tf.placeholder(tf.float32,[None,460,460,3])
lab_holder = tf.placeholder(tf.int32,[None,460,460])
self.net_body = seg_main_body(inp_holder)
seg_layer = self.segmentation_layer(self.net_body.feature_layer,12)
self.build_loss(seg_layer,lab_holder)
self.saver = tf.train.Saver()
self.sess = tf.Session()
M.loadSess('./savings_bgfg/',self.sess,init=True,var_list=M.get_trainable_vars('bg_fg/WideRes'))
self.inp_holder = inp_holder
self.lab_holder = lab_holder
self.seg_layer = seg_layer
def __init__(self, class_num):
self.size = 460
self.class_num = 20
# build placeholders
inp_holder = tf.placeholder(tf.float32, [None, size, size, 3],
name='image_holder')
seg_holder = tf.placeholder(tf.float32, [None, size, size, class_num],
name='segment_holder')
mask_holder = tf.placeholder(tf.float32, [None, size, size],
name='mask_holder')
coord_holder = tf.placeholder(tf.float32, [None, size, size, 6],
name='coordinate_holder')
inst_holder = tf.placeholder(tf.float32, [None, class_num],
name='instance_holder')
# construct input (4 -> 3 with 1x1 conv)
merged_layer = self.merging_layer(inp_holder, seg_holder, mask_holder)
# build network
self.get_coord(size)
self.net_body = seg_main_body(merged_layer)
stream_list = self.get_stream_list(self.net_body.feature_maps)
inst_pred = self.inst_layer(self.net_body.feature_layer,
stream_list[-1], class_num)
self.build_loss(seg_layer, stream_list, inst_pred, lab_holder,
mask_holder, coord_holder, inst_holder)
# build saver and session
self.saver = tf.train.Saver()
self.sess = tf.Session()
# self.writer = tf.summary.FileWriter('./logs/',self.sess.graph)
M.loadSess('./model/',
self.sess,
init=True,
var_list=self.net_body.var)
# set class variables
# holders
self.inp_holder = inp_holder
self.lab_holder = lab_holder
self.mask_holder = mask_holder
self.coord_holder = coord_holder
self.inst_holder = inst_holder
# layers
self.coord_layer = stream_list[-1]
self.inst_layer = inst_pred
def __init__(self):
inp_holder = tf.placeholder(tf.float32, [None, 460, 460, 3])
lab_holder = tf.placeholder(tf.int32, [None, 460, 460])
self.net_body = seg_main_body(inp_holder)
seg_layer = self.segmentation_layer(self.net_body.feature_layer, 12)
self.build_loss(seg_layer, lab_holder)
self.saver = tf.train.Saver()
self.sess = tf.Session()
M.loadSess('./model/',
self.sess,
init=True,
var_list=self.net_body.var)
self.inp_holder = inp_holder
self.lab_holder = lab_holder
self.seg_layer = seg_layer
def evaluate():
import cv2
# listf = open('set1list.txt')
listf = open('1k5_set2.txt')
imglist = []
for i in listf:
imglist.append(i.replace('\n', ''))
with tf.Session() as sess:
M.loadSess(modelpath, sess=sess, modpath=modelpath + modelname)
res = []
imgs = []
print('reading images...')
for pic in imglist:
img = cv2.imread(pic, 1)
img = cv2.resize(img, (122, 144))
M2 = np.float32([[1, 0, 11], [0, 1, 0]])
img = cv2.warpAffine(img, M2, (144, 144))
img = cv2.flip(img, 1)
img = np.float32(img)[8:136, 8:136]
img = cv2.resize(img, (224, 224))
imgs.append(img)
splits = len(imgs) // 100
print('Splits in total...')
for i in range(splits - 1):
cl = sess.run(tf.nn.softmax(evallayer),
feed_dict={imgholder: imgs[i * 100:i * 100 + 100]})
print(cl.shape)
res.append(cl)
cl = sess.run(tf.nn.softmax(evallayer),
feed_dict={imgholder: imgs[(splits - 1) * 100:]})
print(cl.shape)
res.append(cl)
res = np.concatenate(res, axis=0)
print(res.shape)
import scipy.io as sio
sio.savemat('enf56_1.mat', {'data': res})
fout = 'testresult.txt'
fout = open(fout, 'w')
for i in res:
fout.write(str(i) + '\n')
fout.close()
def test():
args = parser.parse_args()
test_acc = 0
rgb_file_txt = '/home/wtx/RGBD_dataset/eaststation/test/test_3Dgallery.txt'
depth_file_txt = '/home/wtx/RGBD_dataset/eaststation/test/test_3Dprobe.txt'
root_folder = '/home/wtx/RGBD_dataset/eaststation/'
imgs, labs = concat_rgb_and_depth(root_folder, rgb_file_txt, depth_file_txt)
img_holder = tf.placeholder(tf.float32, [None, 128, 128, 2])
lab_holder = tf.placeholder(tf.int64, [None])
acc = _LCNN9(img_holder, lab_holder)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
M.loadSess('../tfmodel/', sess)
for iter in range(len(labs) // args.batch_size):
test_acc += sum(sess.run([acc], feed_dict={
img_holder: imgs[iter * args.batch_size:(iter + 1) * args.batch_size],
lab_holder: labs[iter * args.batch_size:(iter + 1) * args.batch_size]}))
sess.close()
ave_acc = test_acc / (len(labs) // args.batch_size)
print('The Accuracy in Test Set:' + str(ave_acc))
def __init__(self,class_num):
img_holder = tf.placeholder(tf.float32,[None,460,460,3])
with tf.variable_scope('bg_fg'):
net_bgfg = network_bg_fg(img_holder)
with tf.variable_scope('seg_part'):
net_seg = network_seg(class_num,img_holder,tf.nn.softmax(tf.image.resize_images(net_bgfg.seg_layer,[460,460]),1)[:,:,:,1])
with tf.variable_scope('inst_part'):
net_inst = network_inst(class_num,img_holder
tf.nn.softmax(tf.image.resize_images(net_bgfg.seg_layer,[460,460]),1)[:,:,:,1],
tf.image.resize_images(net_seg.seg_layer,[460,460]))
self.network_bg_fg = network_bg_fg
self.network_seg = network_seg
self.network_inst = network_inst
self.img_holder = img_holder
self.mask_holder = network_bg_fg.lab_holder
self.seg_holder = network_seg.lab_holder
self.coord_holder = network_inst.coord_holder
self.inst_holder = network_inst.inst_holder
self.mask_out = network_bg_fg.seg_layer
self.seg_out = network_seg.seg_layer
self.inst_num_out = network_inst.inst_layer
self.coord_out = network_inst.coord_layer
self.build_loss()
self.sess = tf.Session()
M.loadSess('./savings_bgfg/',sess=self.sess,init=True,var_list=M.get_all_vars('bg_fg'))
M.loadSess('./savings_seg/',sess=self.sess,var_list=M.get_all_vars('seg_part'))
M.loadSess('./savings_inst/',sess=self.sess,var_list=M.get_all_vars('inst_part'))
def training():
merged = tf.summary.merge_all()
data = getData()
saver = tf.train.Saver()
with tf.Session() as sess:
writer = tf.summary.FileWriter('./logs/', sess.graph)
M.loadSess('./model/', sess=sess)
counter = M.counter
for i in range(1000000):
counter += 1
sample = random.sample(data, BSIZE)
x_train = [i[0] for i in sample]
y_train = [i[1] for i in sample]
a = np.random.uniform(size=[BSIZE, ZDIM], low=-1.0, high=1.0)
a = a / np.linalg.norm(a, axis=1, keepdims=True)
# ge = sess.run(generated,feed_dict={z:a})
for _ in range(5):
sess.run(trainG, feed_dict={z: a, imgholder: x_train})
_, _, mg, lsd, lsg, lsc = sess.run(
[trainC, trainD, merged, lossD, lossG, lossC],
feed_dict={
z: a,
imgholder: x_train,
classholder: y_train
})
if (i) % 5 == 0:
writer.add_summary(mg, counter)
print('iter:', i)
print('lsd:', lsd)
print('lsg:', lsg)
print('lsc:', lsc)
if (i + 1) % 100 == 0:
getGeneratedImg(sess, i + 1)
if (i + 1) % 1000 == 0:
saver.save(sess,
'./model/ModelCounter' + str(counter) + '.ckpt')
def getSample():
with tf.Session() as sess:
data = getData()
M.loadSess('./model/', sess=sess)
for i in range(20):
x_train = random.sample(data, 1)
# print(x_train[0].shape)
x_train = np.float32(x_train[0][0]).reshape([-1, 128, 128, 1])
for j in range(8):
# a = np.random.uniform(size=[1,ZDIM],low=-0.2,high=0.2)
a = np.zeros([1, ZDIM], dtype=np.float32)
genimg = sess.run(generated,
feed_dict={
imgholder: x_train,
noise: a
})
genimg = (genimg + 1) * 127
genimg = genimg.astype(np.uint8)
cv2.imwrite('./sampleimg/' + str(i) + 'gen' + str(j) + '.jpg',
cv2.resize(genimg[0], (128, 128)))
cv2.imwrite(
'./sampleimg/' + str(i) + 'org.jpg',
cv2.resize(((x_train[0] + 1) * 127).astype(np.uint8),
(128, 128)))
def tfreord_train(tfrecord_path):
args = parser.parse_args()
# rgb_file_txt = '/Volumes/Untitled/eaststation/test/test_3Dtexture.txt'
# depth_file_txt = '/Volumes/Untitled/eaststation/test/test_3Ddepth.txt'
# root_folder = '/Volumes/Untitled/eaststation/test/'
images, labels = input(args.batch_size, args.batch_size, tfrecord_path)
loss, acc = LCNN29(images, labels)
l2_loss = tf.losses.get_regularization_loss()
loss += l2_loss
train_op = tf.train.MomentumOptimizer(0.00001, 0.9).minimize(loss)
# train_op = tf.train.AdamOptimizer(0.0001).minimize(loss)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
sess, epoch, step = M.loadSess('../tfmodel_LCNN29/', sess)
saver = tf.train.Saver()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
try:
while not coord.should_stop():
_, loss_value, accuracy = sess.run([train_op, loss, acc])
step += 1
if (args.batch_size * step) % args.samples_num == 0:
epoch += 1
if step % 1 == 0:
print('epoch = %d iter = %d loss = %.2f' % (epoch, step, loss_value))
print('accuracy = %.2f' % accuracy)
if step % 200 == 0:
save_path = '../tfmodel_LCNN29/Epoc_' + str(epoch) + '_' + 'Iter_' + str(step) + '.cpkt'
saver.save(sess, save_path)
save_path2 = save_path + '.meta'
save_path3 = save_path + '.index'
save_path4 = save_path + '.data-00000-of-00001'
save_path5 = '../tfmodel_LCNN29/checkpoint'
shutil.copy(save_path2, save_path2.replace('../tfmodel_LCNN29/', '../backup_LCNN29/'))
shutil.copy(save_path3, save_path3.replace('../tfmodel_LCNN29/', '../backup_LCNN29/'))
shutil.copy(save_path4, save_path4.replace('../tfmodel_LCNN29/', '../backup_LCNN29/'))
shutil.copy(save_path5, save_path5.replace('../tfmodel_LCNN29/', '../backup_LCNN29/'))
except tf.errors.OutOfRangeError:
print('Done training for %d steps' % (step))
finally:
coord.request_stop()
coord.join(threads)
def __init__(self, class_num):
img_holder = tf.placeholder(tf.float32, [None, 460, 460, 3])
with tf.variable_scope('bg_fg'):
net_bgfg = network_bg_fg(img_holder)
with tf.variable_scope('seg_part'):
bg_fg_upsample = tf.nn.softmax(
tf.image.resize_images(net_bgfg.seg_layer,
img_holder.get_shape().as_list()[1:3]),
1)[:, :, :, 1]
print(bg_fg_upsample)
input('pause')
net_seg = network_seg(img_holder, class_num, bg_fg_upsample)
with tf.variable_scope('inst_part'):
net_inst = network_inst(
img_holder, class_num,
tf.nn.softmax(
tf.image.resize_images(net_bgfg.seg_layer,
tf.shape(img_holder)[1:3]),
1)[:, :, :, 1],
tf.image.resize_images(net_seg.seg_layer,
tf.shape(img_holder)[1:3]))
self.net_bgfg = net_bgfg
self.net_seg = net_seg
self.net_inst = net_inst
self.img_holder = img_holder
self.mask_out = tf.image.resize_images(net_bgfg.seg_layer,
tf.shape(img_holder)[1:3])
self.seg_out = tf.image.resize_images(net_seg.seg_layer,
tf.shape(img_holder)[1:3])
self.inst_num_out = net_inst.inst_layer
self.coord_out = net_inst.coord_layer
self.sess = tf.Session()
M.loadSess('./savings_bgfg/',
sess=self.sess,
init=True,
var_list=M.get_all_vars('bg_fg'))
M.loadSess('./savings_seg/',
sess=self.sess,
var_list=M.get_all_vars('seg_part'))
M.loadSess('./savings_inst/',
sess=self.sess,
var_list=M.get_all_vars('inst_part'))
def load(self,path):
M.loadSess(sess=self.sess,modpath=path)
mod.conv_layer(3, 128 * 2, activation=1)
mod.sum(d)
mod.conv_layer(1, 256, activation=1)
mod.conv_layer(3, 256, activation=1)
mod.maxpooling_layer(2, 2) #pool4
mod.flatten()
mod.fcnn_layer(512)
feature_layer = mod.get_current_layer()[0]
return feature_layer, img_holder
with tf.variable_scope('LCNN29'):
feature_layer, img_holder = LCNN29()
#
sess = tf.Session()
model_path = './model/Epoc_49_Iter_663.cpkt'
M.loadSess(model_path, sess)
def eval(img):
res = sess.run(feature_layer, feed_dict={img_holder: img})
return res
def __exit__():
sess.close()
block(mod, 128)
block(mod, 128)
block(mod, 128)
block(mod, 128)
mod.convLayer(1, 256, activation=M.PARAM_MFM)
mod.convLayer(3, 256, activation=M.PARAM_MFM)
mod.maxpoolLayer(2)
mod.flatten()
mod.fcLayer(512)
featurelayer = mod.get_current_layer()
# with tf.variable_scope('enforced_layer'):
# classlayer,evallayer = M.enforcedClassfier(featurelayer,512,labholder,BSIZE,CLASS,dropout=1,enforced=True)
return imgholder, featurelayer
with tf.variable_scope('MainModel'):
imgholder, featurelayer = res_18()
sess = tf.Session()
M.loadSess('./model2/', sess=sess, modpath='./modelres/Epoc0Iter20999.ckpt')
# M.loadSess('./model2/',sess=sess,modpath='./model2/Epoc2Iter29099.ckpt')
def eval(img):
res = sess.run(featurelayer, feed_dict={imgholder: img})
return res
def __exit__():
sess.close()
cropped = item[0]
coord = item[1]
lb = get_lb(coord,crds)
lbs.append([cropped,lb])
# x,y,w,h =
# cv2.rectangle(img,())
return lbs
reader = datareader2.reader(height=240,width=320,scale_range=[0.05,1.2])
b0,b1,c0,c1 = netpart.model_out
start_time = time.time()
MAXITER = 100000
with tf.Session() as sess:
saver = tf.train.Saver()
M.loadSess('./model/',sess,init=True,var_list=M.get_trainable_vars('MSRPN'))
for i in range(MAXITER):
img,coord = reader.get_img()
buff_out = sess.run([b0,b1,c0,c1],feed_dict={netpart.inpholder:[img]})
bs,cs = buff_out[:2],buff_out[2:]
lbs = crop(img,bs,cs,coord)
train_imgs = [k[0] for k in lbs]
train_labs = [k[1] for k in lbs]
# for item in lbs:
# cv2.imshow('ad',item[0])
# print(item[1])
# cv2.waitKey(0)
ls,ac,_ = sess.run([net_veri.loss,net_veri.accuracy,net_veri.ts],
feed_dict={net_veri.inputholder:train_imgs,net_veri.labelholder:train_labs})
if i%10==0:
t2 = time.time()
def __init__(self, model_path='./aim_model_gen/'):
self.model_path = model_path
self.inp_holder = tf.placeholder(tf.float32, [None, 128, 128, 3])
self.age_holder = tf.placeholder(tf.float32, [None, 1])
self.age_holder2 = tf.placeholder(tf.float32, [None, 1])
# get attention A and C
age_expanded = self.expand(self.age_holder, self.inp_holder)
aged_feature = tf.concat([age_expanded, self.inp_holder], -1)
A, C = N.generator_att(aged_feature)
# construct synthesized image
generated = A * C + (1. - A) * self.inp_holder
# get attention A2 and C2
age_expanded2 = self.expand(self.age_holder2, generated)
aged_feature2 = tf.concat([age_expanded2, generated], -1)
A2, C2 = N.generator_att(aged_feature2)
generated2 = A2 * C2 + (1. - A2) * generated
# retrieve tensor for adv2 and ae
adv2, age_pred = N.discriminator(generated)
adv2_real, age_pred_real = N.discriminator(self.inp_holder)
adv2_2, age_pred2 = N.discriminator(generated2)
feat = N.feat_encoder(self.inp_holder)
feat1 = N.feat_encoder(generated)
feat2 = N.feat_encoder(generated2)
self.feat_loss = tf.reduce_mean(
tf.square(feat - feat1) + tf.square(feat - feat2))
self.train_feat = tf.train.AdamOptimizer(0.00001).minimize(
self.feat_loss, var_list=M.get_all_vars('gen_att'))
# get gradient penalty
# gamma1 = tf.random_uniform([],0.0,1.0)
# interp1 = gamma1 * generated + (1. - gamma1) * self.inp_holder
# interp1_y, _ = N.discriminator(interp1, 7)
# grad_p1 = tf.gradients(interp1_y, interp1)[0]
# grad_p1 = tf.sqrt(tf.reduce_sum(tf.square(grad_p1),axis=[1,2,3]))
# grad_p1 = tf.reduce_mean(tf.square(grad_p1 - 1.) * 10.)
# gamma2 = tf.random_uniform([],0.0,1.0)
# interp2 = gamma2 * generated + (1. - gamma2) * self.inp_holder
# interp2_y, _ = N.discriminator(interp2, 7)
# grad_p2 = tf.gradients(interp2_y, interp2)[0]
# grad_p2 = tf.sqrt(tf.reduce_sum(tf.square(grad_p2),axis=[1,2,3]))
# grad_p2 = tf.reduce_mean(tf.square(grad_p2 - 1.) * 10.)
grad_p1 = grad_p2 = 0.
# call loss builder functions
self.mc_loss, self.train_mc = self.build_loss_mc(
generated2, self.inp_holder)
self.adv2_loss_d1, self.adv2_loss_g1, self.train_adv2_1 = self.build_loss_adv2(
adv2, adv2_real, grad_p1)
self.adv2_loss_d2, self.adv2_loss_g2, self.train_adv2_2 = self.build_loss_adv2(
adv2_2, adv2_real, grad_p2)
self.age_cls_loss_dis, self.train_ae_dis = self.build_loss_ae_dis(
age_pred_real, self.age_holder2)
self.age_cls_loss_gen, self.train_ae_gen = self.build_loss_ae_gen(
age_pred, self.age_holder)
self.age_cls_loss_gen2, self.train_ae_gen2 = self.build_loss_ae_gen(
age_pred2, self.age_holder2)
self.loss_A, self.train_A = self.build_loss_A(A)
self.loss_A2, self.train_A2 = self.build_loss_A(A2)
self.update_ops()
self.accuracy = M.accuracy(age_pred_real,
tf.argmax(self.age_holder2, -1))
self.A1_l, self.A2_l = tf.reduce_mean(tf.square(A)), tf.reduce_mean(
tf.square(A2))
self.generated = generated
self.A, self.C = A, C
self.sess = tf.Session()
M.loadSess(model_path, self.sess, init=True)
M.loadSess('./aim_model/',
self.sess,
var_list=M.get_all_vars('encoder'))
self.saver = tf.train.Saver()
def placeholder_train(imgs, labels):
args = parser.parse_args()
with tf.name_scope('img_holder'):
img_holder = tf.placeholder(tf.float32, [args.batch_size, 128, 128, 1])
with tf.name_scope('lab_holder'):
lab_holder = tf.placeholder(tf.int64, [args.batch_size])
val_imgs, val_labs = test_list()
loss, acc = LCNN9(img_holder, lab_holder)
# l2_loss = tf.losses.get_regularization_loss()
# l2_loss = tf.reduce_sum(tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
# total_loss = loss + l2_loss
global_step = tf.Variable(0, trainable=False)
learning_rate = tf.train.exponential_decay(learning_rate=args.lr, global_step=global_step,
decay_steps=20 * args.samples_num / args.batch_size,
# decay_steps = 5000
decay_rate=0.6, staircase=True)
train_op = tf.train.MomentumOptimizer(learning_rate, 0.9).minimize(loss, global_step)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
sess, epoch, step = M.loadSess('../tfmodel/', sess)
saver = tf.train.Saver()
for i in range(args.epoch):
for j in range(args.samples_num // args.batch_size):
images = imgs[j * args.batch_size:(j + 1) * args.batch_size]
labs = labels[j * args.batch_size:(j + 1) * args.batch_size]
_, loss_value, lr, accuracy = sess.run([train_op, loss, learning_rate, acc],
feed_dict={img_holder: images, lab_holder: labs})
step += 1
print('epoch = %d iter = %d loss = %.2f learning_rate = % .6f' % (
epoch, step, loss_value, lr))
print('accuracy = %.2f' % accuracy)
if step % 2000 == 0:
save_path = '../tfmodel/Epoc_' + str(epoch) + '_' + 'Iter_' + str(step) + '.cpkt'
saver.save(sess, save_path)
save_path2 = save_path + '.meta'
save_path3 = save_path + '.index'
save_path4 = save_path + '.data-00000-of-00001'
save_path5 = '../tfmodel/checkpoint'
shutil.copy(save_path2, save_path2.replace('../tfmodel/', '../backup/'))
shutil.copy(save_path3, save_path3.replace('../tfmodel/', '../backup/'))
shutil.copy(save_path4, save_path4.replace('../tfmodel/', '../backup/'))
shutil.copy(save_path5, save_path5.replace('../tfmodel/', '../backup/'))
val_acc = 0
for it in range(len(val_labs) // args.batch_size):
val_acc += sum(sess.run([acc], feed_dict={
img_holder: val_imgs[it * args.batch_size:(it + 1) * args.batch_size],
lab_holder: val_labs[it * args.batch_size:(it + 1) * args.batch_size]}))
val_acc = val_acc / (len(val_labs) // args.batch_size)
print('The Accuracy in Val Set:' + str(val_acc))
# test_acc = 0
#rgb_file_txt = '/home/wtx/RGBD_dataset/eaststation/test/test_3Dgallery.txt'
#depth_file_txt = '/home/wtx/RGBD_dataset/eaststation/test/test_3Dprobe.txt'
#root_folder = '/home/wtx/RGBD_dataset/eaststation/'
#test_imgs, test_labs = concat_rgb_and_depth(root_folder, rgb_file_txt, depth_file_txt)
#for iter in range(len(test_labs) // args.batch_size):
# test_acc += sum(sess.run([acc], feed_dict={
# img_holder: test_imgs[iter * args.batch_size:(iter + 1) * args.batch_size],
# lab_holder: test_labs[iter * args.batch_size:(iter + 1) * args.batch_size]}))
#ave_acc = test_acc / (len(test_labs) // args.batch_size)
#print('The Accuracy in Test Set:' + str(ave_acc))
epoch += 1
M2 = np.float32([[1, 0, 11], [0, 1, 0]])
img = cv2.warpAffine(img, M2, (144, 144))
img = img[8:136, 8:136]
data2.append(img)
print('data2 length:', len(data2))
return data2
build_total_graph()
with tf.Session() as sess:
print('Writing log file...')
saver = tf.train.Saver()
reader = data_reader()
# writer = tf.summary.FileWriter('./logs/',sess.graph)
# data2 = get_data2()
M.loadSess('./model/', sess, init=True)
M.loadSess(modpath='/home/psl/7T/chengyu/modelres/Epoc0Iter20999.ckpt',
sess=sess,
var_list=M.get_trainable_vars('MainModel'))
alltensor = [totalLoss, dis_loss, train_d, train_g, updd, updg]
gentensor = [profHolder, x_fake, gtHolder]
for iteration in range(100000):
train_batch = reader.get_batch(BSIZE)
# p2_batch = random.sample(data2,BSIZE)
p_batch = [i[0] for i in train_batch]
l_batch = [i[1] for i in train_batch]
r_batch = [i[2] for i in train_batch]
n_batch = [i[3] for i in train_batch]
m_batch = [i[4] for i in train_batch]
gt_batch = [i[5] for i in train_batch]
lb_batch = [i[6] for i in train_batch]
coords = list(coords)
if coords:
for i in range(len(coords)):
x, y, w, h, _, cat = coords[i]
cv2.rectangle(buff_img, (x - w // 2, y - h // 2),
(x + w // 2, y + h // 2), (0, 255, 0), 2)
cv2.putText(buff_img, str(cat), (x, y),
cv2.FONT_HERSHEY_SCRIPT_SIMPLEX, 3, (0, 0, 255), 6)
cv2.imshow(name, buff_img)
cv2.waitKey(1)
i = 0
sess = tf.Session()
M.loadSess('./model/', sess, init=False)
def get_coord_from_detection(img, name="image"):
b, c, cat = sess.run([netpart.bias, netpart.conf, netpart.cat],
feed_dict={netpart.inpholder: [img]})
res_bias, res_conf = get_img_coord(img, c, b, cat, 64)
if not res_conf:
result_coords = []
elif len(res_conf) == 1:
result_coords = res_bias
else:
result_coords = non_max_sup(res_bias, res_conf)
draw(img, result_coords, name)
return result_coords
nseimg = (nseimg / 127.5 - 1.0)
gtimg = (gtimg / 127.5 - 1.0)
data.append([faceimg, leimg, reimg, nseimg, mthimg, gtimg])
print(gtimg.max(), gtimg.min())
print(faceimg.max(), faceimg.min())
print('Data length:', len(data))
return data
build_total_graph()
with tf.Session() as sess:
print('Writing log file...')
saver = tf.train.Saver()
# writer = tf.summary.FileWriter('./logs/',sess.graph)
data = get_data()
M.loadSess('./model/', sess)
M.loadSess(modpath='./modelres/Epoc0Iter20999.ckpt',
sess=sess,
var_list=M.get_trainable_vars('MainModel'))
alltensor = [totalLoss, dis_loss, train_d, train_g, updd, updg]
gentensor = [profHolder, x_fake, gtHolder]
for iteration in range(100000):
train_batch = random.sample(data, BSIZE)
p_batch = [i[0] for i in train_batch]
l_batch = [i[1] for i in train_batch]
r_batch = [i[2] for i in train_batch]
n_batch = [i[3] for i in train_batch]
m_batch = [i[4] for i in train_batch]
gt_batch = [i[5] for i in train_batch]
z_batch = np.random.uniform(size=[BSIZE, ZDIM], low=-1.0, high=1.0)
feeddict = {
j = aaa % col
x = int(b[i][j][0]) + j * multip + multip // 2
y = int(b[i][j][1]) + i * multip + multip // 2
w = int(b[i][j][2])
h = int(b[i][j][3])
cv2.rectangle(img, (x - w // 2, y - h // 2), (x + w // 2, y + h // 2),
(0, 255, 0), 2)
cv2.imshow(name, img)
cv2.waitKey(1)
n_minibatches = 4
i = 0
with tf.Session() as sess:
saver = tf.train.Saver()
M.loadSess('drive/Colab/hand_gesture/model/', sess, init=False)
while True:
i += 1
for j in range(n_minibatches):
img, train_dic = reader.get_img()
feed_dict_var = {
netpart.inpholder: [img],
netpart.b_labholder: [train_dic[1]],
netpart.c_labholder: [train_dic[0]],
netpart.cat_labholder: [train_dic[2]]
}
if j == 0:
sess.run(netpart.zero_ops, feed_dict=feed_dict_var)
sess.run(netpart.accum_ops, feed_dict=feed_dict_var)
with tf.name_scope('Optimizer'):
train = tf.train.AdamOptimizer(0.00001).minimize(loss)
# print (train)
#
model_path = 'model/'
log_path = 'log/'
list_train = 'hd5_list_train.txt'
list_val = 'hd5_list_val.txt'
f_log = open('log/log.txt', 'a+')
with tf.Session() as sess:
writer = tf.summary.FileWriter(log_path, sess.graph)
sess, epoc, iters = M.loadSess("model/", sess)
saver = tf.train.Saver()
reader = hd5_reader(list_train, list_val, BSIZE, BSIZE)
ITERS = reader.train_epoc
count = 0
for j in range(EPOC):
for i in range(ITERS):
x_train, y_train_ = reader.train_nextbatch()
global BSIZE
BSIZE = reader.train_bsize
# print ('BSIZE:', BSIZE)
y_train = np.zeros([BSIZE,CLASS],dtype=np.int64)
for index in range(BSIZE):
y_train[index][y_train_[index]] = 1
out, cell = lstm.apply(a_split[i], out, cell)
out_decoded = mod(out)
out_split.append(out_decoded)
out = tf.stack(out_split, 1) # should be in shape [None, 3, 1]
label_holder = tf.placeholder(tf.float32, [None, 6, 1])
# loss = tf.reduce_mean(tf.square(label_holder - out))
loss = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=out, labels=label_holder))
train_op = tf.train.AdamOptimizer(0.01).minimize(loss)
with tf.Session() as sess:
M.loadSess('./model/', sess, init=True)
for i in range(10000):
batch_x, batch_y = getInputVector()
mask = np.ones([1, 6, 1])
ls, _ = sess.run([loss, train_op],
feed_dict={
a: [batch_x],
label_holder: [batch_y]
})
if i % 1000 == 0:
print(ls)
test_x, test_y = getInputVector()
o = sess.run(out, feed_dict={a: [test_x]})
print(test_x)
print(test_y)
loss, acc, img_holder, lab_holder = LCNN29()
with tf.name_scope('Optimizer'):
train = tf.train.AdamOptimizer(0.00001).minimize(loss)
# print (train)
#
model_path = './model/'
log_path = './log/'
list_train = 'hd5_list_train.txt'
list_val = 'hd5_list_val.txt'
f_log = open('./log/log.txt', 'a+')
with tf.Session() as sess:
writer = tf.summary.FileWriter(log_path, sess.graph)
sess, epoc, iters = M.loadSess(model_path, sess)
saver = tf.train.Saver()
reader = hd5_reader(list_train, list_val, BSIZE, BSIZE)
ITERS = reader.train_epoc
count = 0
for j in range(EPOC):
for i in range(ITERS):
x_train, y_train_ = reader.train_nextbatch()
global BSIZE
BSIZE = reader.train_bsize
# print ('BSIZE:', BSIZE)
y_train = np.zeros([BSIZE,CLASS],dtype=np.int64)
for index in range(BSIZE):
y_train[index][y_train_[index]] = 1