mirror of
https://github.com/iperov/DeepFaceLab.git
synced 2025-07-06 04:52:13 -07:00
SAEHD: add 'Blur out mask' and 'Denoise DST faceset' options.
This commit is contained in:
iperov
parent
e41f87e682
commit
8e63666390
1 changed files with 82 additions and 58 deletions
|
|
@ -46,6 +46,8 @@ class SAEHDModel(ModelBase):
|
|||
default_masked_training = self.options['masked_training'] = self.load_or_def_option('masked_training', True)
|
||||
default_eyes_mouth_prio = self.options['eyes_mouth_prio'] = self.load_or_def_option('eyes_mouth_prio', False)
|
||||
default_uniform_yaw = self.options['uniform_yaw'] = self.load_or_def_option('uniform_yaw', False)
|
||||
default_blur_out_mask = self.options['blur_out_mask'] = self.load_or_def_option('blur_out_mask', False)
|
||||
default_dst_denoise = self.options['dst_denoise'] = self.load_or_def_option('dst_denoise', False)
|
||||
|
||||
default_adabelief = self.options['adabelief'] = self.load_or_def_option('adabelief', True)
|
||||
|
||||
|
|
@ -138,6 +140,8 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
|
||||
self.options['eyes_mouth_prio'] = io.input_bool ("Eyes and mouth priority", default_eyes_mouth_prio, help_message='Helps to fix eye problems during training like "alien eyes" and wrong eyes direction. Also makes the detail of the teeth higher.')
|
||||
self.options['uniform_yaw'] = io.input_bool ("Uniform yaw distribution of samples", default_uniform_yaw, help_message='Helps to fix blurry side faces due to small amount of them in the faceset.')
|
||||
self.options['blur_out_mask'] = io.input_bool ("Blur out mask", default_blur_out_mask, help_message='Blurs nearby area outside of applied face mask of training samples. The result is the background near the face is smoothed and less noticeable on swapped face. The exact xseg mask in src and dst faceset is required.')
|
||||
self.options['dst_denoise'] = io.input_bool ("Denoise DST faceset.", default_dst_denoise, help_message='Used in RTM(ReadyToMerge) training with RTM DST faceset. Removes high frequency noise keeping edges. Result is better face syncronization with any face. Can be enabled at any time.')
|
||||
|
||||
default_gan_power = self.options['gan_power'] = self.load_or_def_option('gan_power', 0.0)
|
||||
default_gan_patch_size = self.options['gan_patch_size'] = self.load_or_def_option('gan_patch_size', self.options['resolution'] // 8)
|
||||
|
|
@ -228,6 +232,8 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
random_warp = False if self.pretrain else self.options['random_warp']
|
||||
random_src_flip = self.random_src_flip if not self.pretrain else True
|
||||
random_dst_flip = self.random_dst_flip if not self.pretrain else True
|
||||
blur_out_mask = self.options['blur_out_mask']
|
||||
dst_denoise = self.options['dst_denoise']
|
||||
|
||||
if self.pretrain:
|
||||
self.options_show_override['gan_power'] = 0.0
|
||||
|
|
@ -371,6 +377,22 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
gpu_target_dstm = self.target_dstm[batch_slice,:,:,:]
|
||||
gpu_target_dstm_em = self.target_dstm_em[batch_slice,:,:,:]
|
||||
|
||||
gpu_target_srcm_anti = 1-gpu_target_srcm
|
||||
gpu_target_dstm_anti = 1-gpu_target_dstm
|
||||
|
||||
if blur_out_mask:
|
||||
#gpu_target_src = gpu_target_src*gpu_target_srcm_blur + nn.gaussian_blur(gpu_target_src, resolution // 32)*gpu_target_srcm_anti_blur
|
||||
#gpu_target_dst = gpu_target_dst*gpu_target_dstm_blur + nn.gaussian_blur(gpu_target_dst, resolution // 32)*gpu_target_dstm_anti_blur
|
||||
bg_blur_div = 128
|
||||
|
||||
gpu_target_src = gpu_target_src*gpu_target_srcm + \
|
||||
tf.math.divide_no_nan(nn.gaussian_blur(gpu_target_src*gpu_target_srcm_anti, resolution / bg_blur_div),
|
||||
(1-nn.gaussian_blur(gpu_target_srcm, resolution / bg_blur_div) ) ) * gpu_target_srcm_anti
|
||||
|
||||
gpu_target_dst = gpu_target_dst*gpu_target_dstm + \
|
||||
tf.math.divide_no_nan(nn.gaussian_blur(gpu_target_dst*gpu_target_dstm_anti, resolution / bg_blur_div),
|
||||
(1-nn.gaussian_blur(gpu_target_dstm, resolution / bg_blur_div)) ) * gpu_target_dstm_anti
|
||||
|
||||
# process model tensors
|
||||
if 'df' in archi_type:
|
||||
gpu_src_code = self.inter(self.encoder(gpu_warped_src))
|
||||
|
|
@ -641,7 +663,7 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
SampleGeneratorFace(training_data_src_path, random_ct_samples_path=random_ct_samples_path, debug=self.is_debug(), batch_size=self.get_batch_size(),
|
||||
sample_process_options=SampleProcessor.Options(random_flip=random_src_flip),
|
||||
output_sample_types = [ {'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':random_warp, 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'ct_mode': ct_mode, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'ct_mode': ct_mode, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'ct_mode': ct_mode, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_MASK, 'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.G, 'face_mask_type' : SampleProcessor.FaceMaskType.FULL_FACE, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_MASK, 'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.G, 'face_mask_type' : SampleProcessor.FaceMaskType.EYES_MOUTH, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
],
|
||||
|
|
@ -651,7 +673,8 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
SampleGeneratorFace(training_data_dst_path, debug=self.is_debug(), batch_size=self.get_batch_size(),
|
||||
sample_process_options=SampleProcessor.Options(random_flip=random_dst_flip),
|
||||
output_sample_types = [ {'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':random_warp, 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'denoise_filter' : dst_denoise, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_MASK, 'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.G, 'face_mask_type' : SampleProcessor.FaceMaskType.FULL_FACE, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
{'sample_type': SampleProcessor.SampleType.FACE_MASK, 'warp':False , 'transform':True, 'channel_type' : SampleProcessor.ChannelType.G, 'face_mask_type' : SampleProcessor.FaceMaskType.EYES_MOUTH, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
|
||||
],
|
||||
|
|
@ -739,27 +762,28 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
bs = self.get_batch_size()
|
||||
|
||||
( (warped_src, target_src, target_srcm, target_srcm_em), \
|
||||
(warped_dst, target_dst, target_dstm, target_dstm_em) ) = self.generate_next_samples()
|
||||
(warped_dst, target_dst, target_dst_train, target_dstm, target_dstm_em) ) = self.generate_next_samples()
|
||||
|
||||
src_loss, dst_loss = self.src_dst_train (warped_src, target_src, target_srcm, target_srcm_em, warped_dst, target_dst, target_dstm, target_dstm_em)
|
||||
src_loss, dst_loss = self.src_dst_train (warped_src, target_src, target_srcm, target_srcm_em, warped_dst, target_dst_train, target_dstm, target_dstm_em)
|
||||
|
||||
for i in range(bs):
|
||||
self.last_src_samples_loss.append ( (target_src[i], target_srcm[i], target_srcm_em[i], src_loss[i] ) )
|
||||
self.last_dst_samples_loss.append ( (target_dst[i], target_dstm[i], target_dstm_em[i], dst_loss[i] ) )
|
||||
self.last_src_samples_loss.append ( (src_loss[i], target_src[i], target_srcm[i], target_srcm_em[i],) )
|
||||
self.last_dst_samples_loss.append ( (dst_loss[i], target_dst[i], target_dst_train[i], target_dstm[i], target_dstm_em[i],) )
|
||||
|
||||
if len(self.last_src_samples_loss) >= bs*16:
|
||||
src_samples_loss = sorted(self.last_src_samples_loss, key=operator.itemgetter(3), reverse=True)
|
||||
dst_samples_loss = sorted(self.last_dst_samples_loss, key=operator.itemgetter(3), reverse=True)
|
||||
src_samples_loss = sorted(self.last_src_samples_loss, key=operator.itemgetter(0), reverse=True)
|
||||
dst_samples_loss = sorted(self.last_dst_samples_loss, key=operator.itemgetter(0), reverse=True)
|
||||
|
||||
target_src = np.stack( [ x[0] for x in src_samples_loss[:bs] ] )
|
||||
target_srcm = np.stack( [ x[1] for x in src_samples_loss[:bs] ] )
|
||||
target_srcm_em = np.stack( [ x[2] for x in src_samples_loss[:bs] ] )
|
||||
target_src = np.stack( [ x[1] for x in src_samples_loss[:bs] ] )
|
||||
target_srcm = np.stack( [ x[2] for x in src_samples_loss[:bs] ] )
|
||||
target_srcm_em = np.stack( [ x[3] for x in src_samples_loss[:bs] ] )
|
||||
|
||||
target_dst = np.stack( [ x[0] for x in dst_samples_loss[:bs] ] )
|
||||
target_dstm = np.stack( [ x[1] for x in dst_samples_loss[:bs] ] )
|
||||
target_dstm_em = np.stack( [ x[2] for x in dst_samples_loss[:bs] ] )
|
||||
target_dst = np.stack( [ x[1] for x in dst_samples_loss[:bs] ] )
|
||||
target_dst_train = np.stack( [ x[2] for x in dst_samples_loss[:bs] ] )
|
||||
target_dstm = np.stack( [ x[3] for x in dst_samples_loss[:bs] ] )
|
||||
target_dstm_em = np.stack( [ x[4] for x in dst_samples_loss[:bs] ] )
|
||||
|
||||
src_loss, dst_loss = self.src_dst_train (target_src, target_src, target_srcm, target_srcm_em, target_dst, target_dst, target_dstm, target_dstm_em)
|
||||
src_loss, dst_loss = self.src_dst_train (target_src, target_src, target_srcm, target_srcm_em, target_dst, target_dst_train, target_dstm, target_dstm_em)
|
||||
self.last_src_samples_loss = []
|
||||
self.last_dst_samples_loss = []
|
||||
|
||||
|
|
@ -767,14 +791,14 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
|
|||
self.D_train (warped_src, warped_dst)
|
||||
|
||||
if self.gan_power != 0:
|
||||
self.D_src_dst_train (warped_src, target_src, target_srcm, target_srcm_em, warped_dst, target_dst, target_dstm, target_dstm_em)
|
||||
self.D_src_dst_train (warped_src, target_src, target_srcm, target_srcm_em, warped_dst, target_dst_train, target_dstm, target_dstm_em)
|
||||
|
||||
return ( ('src_loss', np.mean(src_loss) ), ('dst_loss', np.mean(dst_loss) ), )
|
||||
|
||||
#override
|
||||
def onGetPreview(self, samples, for_history=False):
|
||||
( (warped_src, target_src, target_srcm, target_srcm_em),
|
||||
(warped_dst, target_dst, target_dstm, target_dstm_em) ) = samples
|
||||
(warped_dst, target_dst, target_dst_train, target_dstm, target_dstm_em) ) = samples
|
||||
|
||||
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] ]
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue