mirror of
https://github.com/iperov/DeepFaceLab.git
synced 2025-07-06 04:52:13 -07:00
Upgraded to TF version 1.13.2
Removed the wait at first launch for most graphics cards. Increased speed of training by 10-20%, but you have to retrain all models from scratch. SAEHD: added option 'use float16' Experimental option. Reduces the model size by half. Increases the speed of training. Decreases the accuracy of the model. The model may collapse or not train. Model may not learn the mask in large resolutions. true_face_training option is replaced by "True face power". 0.0000 .. 1.0 Experimental option. Discriminates the result face to be more like the src face. Higher value - stronger discrimination. Comparison - https://i.imgur.com/czScS9q.png
This commit is contained in:
Colombo
parent
a3dfcb91b9
commit
76ca79216e
49 changed files with 1320 additions and 1297 deletions
|
|
@ -13,11 +13,13 @@ from samplelib import *
|
|||
class QModel(ModelBase):
|
||||
#override
|
||||
def on_initialize(self):
|
||||
nn.initialize()
|
||||
device_config = nn.getCurrentDeviceConfig()
|
||||
self.model_data_format = "NCHW" if len(device_config.devices) != 0 else "NHWC"
|
||||
nn.initialize(data_format=self.model_data_format)
|
||||
tf = nn.tf
|
||||
|
||||
conv_kernel_initializer = nn.initializers.ca
|
||||
|
||||
conv_kernel_initializer = nn.initializers.ca()
|
||||
|
||||
class Downscale(nn.ModelBase):
|
||||
def __init__(self, in_ch, out_ch, kernel_size=5, dilations=1, subpixel=True, use_activator=True, *kwargs ):
|
||||
self.in_ch = in_ch
|
||||
|
|
@ -39,7 +41,7 @@ class QModel(ModelBase):
|
|||
x = self.conv1(x)
|
||||
|
||||
if self.subpixel:
|
||||
x = tf.nn.space_to_depth(x, 2)
|
||||
x = nn.tf_space_to_depth(x, 2)
|
||||
|
||||
if self.use_activator:
|
||||
x = nn.tf_gelu(x)
|
||||
|
|
@ -63,7 +65,7 @@ class QModel(ModelBase):
|
|||
for down in self.downs:
|
||||
x = down(x)
|
||||
return x
|
||||
|
||||
|
||||
class Upscale(nn.ModelBase):
|
||||
def on_build(self, in_ch, out_ch, kernel_size=3 ):
|
||||
self.conv1 = nn.Conv2D( in_ch, out_ch*4, kernel_size=kernel_size, padding='SAME', kernel_initializer=conv_kernel_initializer)
|
||||
|
|
@ -71,9 +73,9 @@ class QModel(ModelBase):
|
|||
def forward(self, x):
|
||||
x = self.conv1(x)
|
||||
x = nn.tf_gelu(x)
|
||||
x = tf.nn.depth_to_space(x, 2)
|
||||
x = nn.tf_depth_to_space(x, 2)
|
||||
return x
|
||||
|
||||
|
||||
class ResidualBlock(nn.ModelBase):
|
||||
def on_build(self, ch, kernel_size=3 ):
|
||||
self.conv1 = nn.Conv2D( ch, ch, kernel_size=kernel_size, padding='SAME', kernel_initializer=conv_kernel_initializer)
|
||||
|
|
@ -109,7 +111,7 @@ class QModel(ModelBase):
|
|||
def forward(self, inp):
|
||||
x = self.dense1(inp)
|
||||
x = self.dense2(x)
|
||||
x = tf.reshape (x, (-1, lowest_dense_res, lowest_dense_res, self.ae_out_ch))
|
||||
x = nn.tf_reshape_4D (x, lowest_dense_res, lowest_dense_res, self.ae_out_ch)
|
||||
x = self.upscale1(x)
|
||||
x = self.res1(x)
|
||||
return x
|
||||
|
|
@ -118,11 +120,11 @@ class QModel(ModelBase):
|
|||
return self.ae_out_ch
|
||||
|
||||
class Decoder(nn.ModelBase):
|
||||
def on_build(self, in_ch, d_ch):
|
||||
self.upscale1 = Upscale(in_ch, d_ch*4)
|
||||
self.res1 = ResidualBlock(d_ch*4)
|
||||
self.upscale2 = Upscale(d_ch*4, d_ch*2)
|
||||
self.res2 = ResidualBlock(d_ch*2)
|
||||
def on_build(self, in_ch, d_ch):
|
||||
self.upscale1 = Upscale(in_ch, d_ch*4)
|
||||
self.res1 = ResidualBlock(d_ch*4)
|
||||
self.upscale2 = Upscale(d_ch*4, d_ch*2)
|
||||
self.res2 = ResidualBlock(d_ch*2)
|
||||
self.upscale3 = Upscale(d_ch*2, d_ch*1)
|
||||
self.res3 = ResidualBlock(d_ch*1)
|
||||
|
||||
|
|
@ -134,8 +136,8 @@ class QModel(ModelBase):
|
|||
self.out_convm = nn.Conv2D( d_ch//2, 1, kernel_size=1, padding='SAME', kernel_initializer=conv_kernel_initializer)
|
||||
|
||||
def forward(self, inp):
|
||||
z = inp
|
||||
x = self.upscale1 (z)
|
||||
z = inp
|
||||
x = self.upscale1 (z)
|
||||
x = self.res1 (x)
|
||||
x = self.upscale2 (x)
|
||||
x = self.res2 (x)
|
||||
|
|
@ -158,7 +160,7 @@ class QModel(ModelBase):
|
|||
d_dims = 64
|
||||
self.pretrain = False
|
||||
self.pretrain_just_disabled = False
|
||||
|
||||
|
||||
masked_training = True
|
||||
|
||||
models_opt_on_gpu = len(devices) == 1 and devices[0].total_mem_gb >= 4
|
||||
|
|
@ -167,8 +169,8 @@ class QModel(ModelBase):
|
|||
|
||||
input_nc = 3
|
||||
output_nc = 3
|
||||
bgr_shape = (resolution, resolution, output_nc)
|
||||
mask_shape = (resolution, resolution, 1)
|
||||
bgr_shape = nn.get4Dshape(resolution,resolution,input_nc)
|
||||
mask_shape = nn.get4Dshape(resolution,resolution,1)
|
||||
lowest_dense_res = resolution // 16
|
||||
|
||||
self.model_filename_list = []
|
||||
|
|
@ -176,22 +178,22 @@ class QModel(ModelBase):
|
|||
|
||||
with tf.device ('/CPU:0'):
|
||||
#Place holders on CPU
|
||||
self.warped_src = tf.placeholder (tf.float32, (None,)+bgr_shape)
|
||||
self.warped_dst = tf.placeholder (tf.float32, (None,)+bgr_shape)
|
||||
self.warped_src = tf.placeholder (nn.tf_floatx, bgr_shape)
|
||||
self.warped_dst = tf.placeholder (nn.tf_floatx, bgr_shape)
|
||||
|
||||
self.target_src = tf.placeholder (tf.float32, (None,)+bgr_shape)
|
||||
self.target_dst = tf.placeholder (tf.float32, (None,)+bgr_shape)
|
||||
self.target_src = tf.placeholder (nn.tf_floatx, bgr_shape)
|
||||
self.target_dst = tf.placeholder (nn.tf_floatx, bgr_shape)
|
||||
|
||||
self.target_srcm = tf.placeholder (tf.float32, (None,)+mask_shape)
|
||||
self.target_dstm = tf.placeholder (tf.float32, (None,)+mask_shape)
|
||||
self.target_srcm = tf.placeholder (nn.tf_floatx, mask_shape)
|
||||
self.target_dstm = tf.placeholder (nn.tf_floatx, mask_shape)
|
||||
|
||||
# Initializing model classes
|
||||
with tf.device (models_opt_device):
|
||||
self.encoder = Encoder(in_ch=input_nc, e_ch=e_dims, name='encoder')
|
||||
encoder_out_ch = self.encoder.compute_output_shape ( (tf.float32, (None,resolution,resolution,input_nc)))[-1]
|
||||
encoder_out_ch = self.encoder.compute_output_channels ( (nn.tf_floatx, bgr_shape))
|
||||
|
||||
self.inter = Inter (in_ch=encoder_out_ch, lowest_dense_res=lowest_dense_res, ae_ch=ae_dims, ae_out_ch=ae_dims, d_ch=d_dims, name='inter')
|
||||
inter_out_ch = self.inter.compute_output_shape ( (tf.float32, (None,encoder_out_ch)))[-1]
|
||||
inter_out_ch = self.inter.compute_output_channels ( (nn.tf_floatx, (None,encoder_out_ch)))
|
||||
|
||||
self.decoder_src = Decoder(in_ch=inter_out_ch, d_ch=d_dims, name='decoder_src')
|
||||
self.decoder_dst = Decoder(in_ch=inter_out_ch, d_ch=d_dims, name='decoder_dst')
|
||||
|
|
@ -203,7 +205,7 @@ class QModel(ModelBase):
|
|||
|
||||
if self.is_training:
|
||||
self.src_dst_trainable_weights = self.encoder.get_weights() + self.inter.get_weights() + self.decoder_src.get_weights() + self.decoder_dst.get_weights()
|
||||
|
||||
|
||||
# Initialize optimizers
|
||||
self.src_dst_opt = nn.TFRMSpropOptimizer(lr=2e-4, lr_dropout=0.3, name='src_dst_opt')
|
||||
self.src_dst_opt.initialize_variables(self.src_dst_trainable_weights, vars_on_cpu=optimizer_vars_on_cpu )
|
||||
|
|
@ -222,7 +224,7 @@ class QModel(ModelBase):
|
|||
gpu_pred_src_srcm_list = []
|
||||
gpu_pred_dst_dstm_list = []
|
||||
gpu_pred_src_dstm_list = []
|
||||
|
||||
|
||||
gpu_src_losses = []
|
||||
gpu_dst_losses = []
|
||||
gpu_src_dst_loss_gvs = []
|
||||
|
|
@ -239,7 +241,7 @@ class QModel(ModelBase):
|
|||
gpu_target_srcm = self.target_srcm[batch_slice,:,:,:]
|
||||
gpu_target_dstm = self.target_dstm[batch_slice,:,:,:]
|
||||
|
||||
# process model tensors
|
||||
# process model tensors
|
||||
gpu_src_code = self.inter(self.encoder(gpu_warped_src))
|
||||
gpu_dst_code = self.inter(self.encoder(gpu_warped_dst))
|
||||
gpu_pred_src_src, gpu_pred_src_srcm = self.decoder_src(gpu_src_code)
|
||||
|
|
@ -249,11 +251,11 @@ class QModel(ModelBase):
|
|||
gpu_pred_src_src_list.append(gpu_pred_src_src)
|
||||
gpu_pred_dst_dst_list.append(gpu_pred_dst_dst)
|
||||
gpu_pred_src_dst_list.append(gpu_pred_src_dst)
|
||||
|
||||
|
||||
gpu_pred_src_srcm_list.append(gpu_pred_src_srcm)
|
||||
gpu_pred_dst_dstm_list.append(gpu_pred_dst_dstm)
|
||||
gpu_pred_src_dstm_list.append(gpu_pred_src_dstm)
|
||||
|
||||
|
||||
gpu_target_srcm_blur = nn.tf_gaussian_blur(gpu_target_srcm, max(1, resolution // 32) )
|
||||
gpu_target_dstm_blur = nn.tf_gaussian_blur(gpu_target_dstm, max(1, resolution // 32) )
|
||||
|
||||
|
|
@ -271,11 +273,11 @@ class QModel(ModelBase):
|
|||
|
||||
gpu_src_loss = tf.reduce_mean ( 10*nn.tf_dssim(gpu_target_srcmasked_opt, gpu_pred_src_src_masked_opt, max_val=1.0, filter_size=int(resolution/11.6)), axis=[1])
|
||||
gpu_src_loss += tf.reduce_mean ( 10*tf.square ( gpu_target_srcmasked_opt - gpu_pred_src_src_masked_opt ), axis=[1,2,3])
|
||||
gpu_src_loss += tf.reduce_mean ( tf.square( gpu_target_srcm - gpu_pred_src_srcm ),axis=[1,2,3] )
|
||||
gpu_src_loss += tf.reduce_mean ( 10*tf.square( gpu_target_srcm - gpu_pred_src_srcm ),axis=[1,2,3] )
|
||||
|
||||
gpu_dst_loss = tf.reduce_mean ( 10*nn.tf_dssim(gpu_target_dst_masked_opt, gpu_pred_dst_dst_masked_opt, max_val=1.0, filter_size=int(resolution/11.6) ), axis=[1])
|
||||
gpu_dst_loss += tf.reduce_mean ( 10*tf.square( gpu_target_dst_masked_opt- gpu_pred_dst_dst_masked_opt ), axis=[1,2,3])
|
||||
gpu_dst_loss += tf.reduce_mean ( tf.square( gpu_target_dstm - gpu_pred_dst_dstm ),axis=[1,2,3] )
|
||||
gpu_dst_loss += tf.reduce_mean ( 10*tf.square( gpu_target_dstm - gpu_pred_dst_dstm ),axis=[1,2,3] )
|
||||
|
||||
gpu_src_losses += [gpu_src_loss]
|
||||
gpu_dst_losses += [gpu_dst_loss]
|
||||
|
|
@ -286,29 +288,16 @@ class QModel(ModelBase):
|
|||
|
||||
# Average losses and gradients, and create optimizer update ops
|
||||
with tf.device (models_opt_device):
|
||||
if gpu_count == 1:
|
||||
pred_src_src = gpu_pred_src_src_list[0]
|
||||
pred_dst_dst = gpu_pred_dst_dst_list[0]
|
||||
pred_src_dst = gpu_pred_src_dst_list[0]
|
||||
pred_src_srcm = gpu_pred_src_srcm_list[0]
|
||||
pred_dst_dstm = gpu_pred_dst_dstm_list[0]
|
||||
pred_src_dstm = gpu_pred_src_dstm_list[0]
|
||||
|
||||
src_loss = gpu_src_losses[0]
|
||||
dst_loss = gpu_dst_losses[0]
|
||||
src_dst_loss_gv = gpu_src_dst_loss_gvs[0]
|
||||
else:
|
||||
pred_src_src = tf.concat(gpu_pred_src_src_list, 0)
|
||||
pred_dst_dst = tf.concat(gpu_pred_dst_dst_list, 0)
|
||||
pred_src_dst = tf.concat(gpu_pred_src_dst_list, 0)
|
||||
pred_src_srcm = tf.concat(gpu_pred_src_srcm_list, 0)
|
||||
pred_dst_dstm = tf.concat(gpu_pred_dst_dstm_list, 0)
|
||||
pred_src_dstm = tf.concat(gpu_pred_src_dstm_list, 0)
|
||||
|
||||
src_loss = nn.tf_average_tensor_list(gpu_src_losses)
|
||||
dst_loss = nn.tf_average_tensor_list(gpu_dst_losses)
|
||||
src_dst_loss_gv = nn.tf_average_gv_list (gpu_src_dst_loss_gvs)
|
||||
pred_src_src = nn.tf_concat(gpu_pred_src_src_list, 0)
|
||||
pred_dst_dst = nn.tf_concat(gpu_pred_dst_dst_list, 0)
|
||||
pred_src_dst = nn.tf_concat(gpu_pred_src_dst_list, 0)
|
||||
pred_src_srcm = nn.tf_concat(gpu_pred_src_srcm_list, 0)
|
||||
pred_dst_dstm = nn.tf_concat(gpu_pred_dst_dstm_list, 0)
|
||||
pred_src_dstm = nn.tf_concat(gpu_pred_src_dstm_list, 0)
|
||||
|
||||
src_loss = nn.tf_average_tensor_list(gpu_src_losses)
|
||||
dst_loss = nn.tf_average_tensor_list(gpu_dst_losses)
|
||||
src_dst_loss_gv = nn.tf_average_gv_list (gpu_src_dst_loss_gvs)
|
||||
src_dst_loss_gv_op = self.src_dst_opt.get_update_op (src_dst_loss_gv)
|
||||
|
||||
# Initializing training and view functions
|
||||
|
|
@ -341,17 +330,15 @@ class QModel(ModelBase):
|
|||
_, gpu_pred_dst_dstm = self.decoder_dst(gpu_dst_code)
|
||||
|
||||
def AE_merge( warped_dst):
|
||||
|
||||
return nn.tf_sess.run ( [gpu_pred_src_dst, gpu_pred_dst_dstm, gpu_pred_src_dstm], feed_dict={self.warped_dst:warped_dst})
|
||||
|
||||
self.AE_merge = AE_merge
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
# Loading/initializing all models/optimizers weights
|
||||
for model, filename in io.progress_bar_generator(self.model_filename_list, "Initializing models"):
|
||||
do_init = self.is_first_run()
|
||||
|
||||
|
||||
if self.pretrain_just_disabled:
|
||||
if model == self.inter:
|
||||
do_init = True
|
||||
|
|
@ -359,16 +346,15 @@ class QModel(ModelBase):
|
|||
if not do_init:
|
||||
do_init = not model.load_weights( self.get_strpath_storage_for_file(filename) )
|
||||
|
||||
if do_init and self.pretrained_model_path is not None:
|
||||
if do_init and self.pretrained_model_path is not None:
|
||||
pretrained_filepath = self.pretrained_model_path / filename
|
||||
if pretrained_filepath.exists():
|
||||
do_init = not model.load_weights(pretrained_filepath)
|
||||
|
||||
|
||||
if do_init:
|
||||
model.init_weights()
|
||||
|
||||
# initializing sample generators
|
||||
|
||||
if self.is_training:
|
||||
t = SampleProcessor.Types
|
||||
face_type = t.FACE_TYPE_FULL
|
||||
|
|
@ -384,19 +370,19 @@ class QModel(ModelBase):
|
|||
self.set_training_data_generators ([
|
||||
SampleGeneratorFace(training_data_src_path, debug=self.is_debug(), batch_size=self.get_batch_size(),
|
||||
sample_process_options=SampleProcessor.Options(random_flip=True if self.pretrain else False),
|
||||
output_sample_types = [ {'types' : (t.IMG_WARPED_TRANSFORMED, face_type, t.MODE_BGR), 'resolution':resolution, },
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_BGR), 'resolution': resolution, },
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_M), 'resolution': resolution } ],
|
||||
output_sample_types = [ {'types' : (t.IMG_WARPED_TRANSFORMED, face_type, t.MODE_BGR), 'data_format':nn.data_format, 'resolution':resolution, },
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_BGR), 'data_format':nn.data_format, 'resolution': resolution, },
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_M), 'data_format':nn.data_format, 'resolution': resolution } ],
|
||||
generators_count=src_generators_count ),
|
||||
|
||||
SampleGeneratorFace(training_data_dst_path, debug=self.is_debug(), batch_size=self.get_batch_size(),
|
||||
sample_process_options=SampleProcessor.Options(random_flip=True if self.pretrain else False),
|
||||
output_sample_types = [ {'types' : (t.IMG_WARPED_TRANSFORMED, face_type, t.MODE_BGR), 'resolution':resolution},
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_BGR), 'resolution': resolution},
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_M), 'resolution': resolution} ],
|
||||
output_sample_types = [ {'types' : (t.IMG_WARPED_TRANSFORMED, face_type, t.MODE_BGR), 'data_format':nn.data_format, 'resolution':resolution},
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_BGR), 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'types' : (t.IMG_TRANSFORMED, face_type, t.MODE_M), 'data_format':nn.data_format, 'resolution': resolution} ],
|
||||
generators_count=dst_generators_count )
|
||||
])
|
||||
|
||||
|
||||
self.last_samples = None
|
||||
|
||||
#override
|
||||
|
|
@ -408,22 +394,21 @@ class QModel(ModelBase):
|
|||
for model, filename in io.progress_bar_generator(self.get_model_filename_list(), "Saving", leave=False):
|
||||
model.save_weights ( self.get_strpath_storage_for_file(filename) )
|
||||
|
||||
|
||||
#override
|
||||
def onTrainOneIter(self):
|
||||
if self.get_iter() % 3 == 0 and self.last_samples is not None:
|
||||
( (warped_src, target_src, target_srcm), \
|
||||
(warped_dst, target_dst, target_dstm) ) = self.last_samples
|
||||
src_loss, dst_loss = self.src_dst_train (target_src, target_src, target_srcm,
|
||||
(warped_dst, target_dst, target_dstm) ) = self.last_samples
|
||||
src_loss, dst_loss = self.src_dst_train (target_src, target_src, target_srcm,
|
||||
target_dst, target_dst, target_dstm)
|
||||
else:
|
||||
samples = self.last_samples = self.generate_next_samples()
|
||||
( (warped_src, target_src, target_srcm), \
|
||||
(warped_dst, target_dst, target_dstm) ) = samples
|
||||
|
||||
src_loss, dst_loss = self.src_dst_train (warped_src, target_src, target_srcm,
|
||||
src_loss, dst_loss = self.src_dst_train (warped_src, target_src, target_srcm,
|
||||
warped_dst, target_dst, target_dstm)
|
||||
|
||||
|
||||
return ( ('src_loss', src_loss), ('dst_loss', dst_loss), )
|
||||
|
||||
#override
|
||||
|
|
@ -435,9 +420,11 @@ class QModel(ModelBase):
|
|||
[ [sample[0:n_samples] for sample in sample_list ]
|
||||
for sample_list in samples ]
|
||||
|
||||
S, D, SS, DD, DDM, SD, SDM = [ np.clip(x, 0.0, 1.0) for x in ([target_src,target_dst] + self.AE_view (target_src, target_dst) ) ]
|
||||
S, D, SS, DD, DDM, SD, SDM = [ np.clip( nn.to_data_format(x,"NHWC", self.model_data_format), 0.0, 1.0) for x in ([target_src,target_dst] + self.AE_view (target_src, target_dst) ) ]
|
||||
DDM, SDM, = [ np.repeat (x, (3,), -1) for x in [DDM, SDM] ]
|
||||
|
||||
target_srcm, target_dstm = [ nn.to_data_format(x,"NHWC", self.model_data_format) for x in ([target_srcm, target_dstm] )]
|
||||
|
||||
result = []
|
||||
st = []
|
||||
for i in range(n_samples):
|
||||
|
|
@ -456,8 +443,10 @@ class QModel(ModelBase):
|
|||
return result
|
||||
|
||||
def predictor_func (self, face=None):
|
||||
face = face[None,...]
|
||||
face = nn.to_data_format(face, self.model_data_format, "NHWC")
|
||||
|
||||
bgr, mask_dst_dstm, mask_src_dstm = self.AE_merge (face[np.newaxis,...])
|
||||
bgr, mask_dst_dstm, mask_src_dstm = [ nn.to_data_format(x, "NHWC", self.model_data_format).astype(np.float32) for x in self.AE_merge (face) ]
|
||||
mask = mask_dst_dstm[0] * mask_src_dstm[0]
|
||||
return bgr[0], mask[...,0]
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue