github/DeepFaceLab
1
0
Fork 0
mirror of https://github.com/iperov/DeepFaceLab.git synced 2025-08-21 05:53:24 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions

fs augmentation

Browse source
This commit is contained in:
sinofis 2020-10-14 15:19:23 +02:00
commit 97295c2246
4 changed files with 64 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

2
core/imagelib/__init__.py
View file

@ -12,7 +12,7 @@ from .warp import gen_warp_params, warp_by_params
from .reduce_colors import reduce_colors
from .color_transfer import color_transfer, color_transfer_mix, color_transfer_sot, color_transfer_mkl, color_transfer_idt, color_hist_match, reinhard_color_transfer, linear_color_transfer
from .color_transfer import color_transfer, color_transfer_mix, color_transfer_sot, color_transfer_mkl, color_transfer_idt, color_hist_match, reinhard_color_transfer, linear_color_transfer, color_augmentation
from .common import normalize_channels, cut_odd_image, overlay_alpha_image

49
core/imagelib/color_transfer.py
View file

@ -366,3 +366,52 @@ def color_transfer(ct_mode, img_src, img_trg):
else:
raise ValueError(f"unknown ct_mode {ct_mode}")
return out
# imported from faceswap
def color_augmentation(img):
""" Color adjust RGB image """
face = img
face = (face * 255.0).astype("uint8")
face = random_clahe(face)
face = random_lab(face)
img[:, :, :3] = face
return img.astype('float32') / 255.0
def random_lab(image):
""" Perform random color/lightness adjustment in L*a*b* colorspace """
amount_l = 30 / 100
amount_ab = 8 / 100
randoms = [(random() * amount_l * 2) - amount_l, # L adjust
(random() * amount_ab * 2) - amount_ab, # A adjust
(random() * amount_ab * 2) - amount_ab] # B adjust
image = cv2.cvtColor( # pylint:disable=no-member
image, cv2.COLOR_BGR2LAB).astype("float32") / 255.0 # pylint:disable=no-member
for idx, adjustment in enumerate(randoms):
if adjustment >= 0:
image[:, :, idx] = ((1 - image[:, :, idx]) * adjustment) + image[:, :, idx]
else:
image[:, :, idx] = image[:, :, idx] * (1 + adjustment)
image = cv2.cvtColor((image * 255.0).astype("uint8"), # pylint:disable=no-member
cv2.COLOR_LAB2BGR) # pylint:disable=no-member
return image
def random_clahe(image):
""" Randomly perform Contrast Limited Adaptive Histogram Equalization """
contrast_random = random()
if contrast_random > 50 / 100:
return image
# base_contrast = image.shape[0] // 128
base_contrast = 1 # testing because it breaks on small sizes
grid_base = random() * 4
contrast_adjustment = int(grid_base * (base_contrast / 2))
grid_size = base_contrast + contrast_adjustment
clahe = cv2.createCLAHE(clipLimit=2.0, # pylint: disable=no-member
tileGridSize=(grid_size, grid_size))
for chan in range(3):
image[:, :, chan] = clahe.apply(image[:, :, chan])
return image

9
models/Model_SAEHD/Model.py
View file

@ -152,7 +152,7 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
self.options['face_style_power'] = np.clip ( io.input_number("Face style power", default_face_style_power, add_info="0.0..100.0", help_message="Learn the color of the predicted face to be the same as dst inside mask. If you want to use this option with 'whole_face' you have to use XSeg trained mask. Warning: Enable it only after 10k iters, when predicted face is clear enough to start learn style. Start from 0.001 value and check history changes. Enabling this option increases the chance of model collapse."), 0.0, 100.0 )
self.options['bg_style_power'] = np.clip ( io.input_number("Background style power", default_bg_style_power, add_info="0.0..100.0", help_message="Learn the area outside mask of the predicted face to be the same as dst. If you want to use this option with 'whole_face' you have to use XSeg trained mask. For whole_face you have to use XSeg trained mask. This can make face more like dst. Enabling this option increases the chance of model collapse. Typical value is 2.0"), 0.0, 100.0 )
self.options['ct_mode'] = io.input_str (f"Color transfer for src faceset", default_ct_mode, ['none','rct','lct','mkl','idt','sot'], help_message="Change color distribution of src samples close to dst samples. Try all modes to find the best.")
self.options['ct_mode'] = io.input_str (f"Color transfer for src faceset", default_ct_mode, ['none','rct','lct','mkl','idt','sot', 'fs-aug'], help_message="Change color distribution of src samples close to dst samples. Try all modes to find the best. FS aug adds random color to dst and src")
self.options['clipgrad'] = io.input_bool ("Enable gradient clipping", default_clipgrad, help_message="Gradient clipping reduces chance of model collapse, sacrificing speed of training.")
self.options['pretrain'] = io.input_bool ("Enable pretraining mode", default_pretrain, help_message="Pretrain the model with large amount of various faces. After that, model can be used to train the fakes more quickly.")
@ -615,10 +615,13 @@ Examples: df, liae, df-d, df-ud, liae-ud, ...
uniform_yaw_distribution=self.options['uniform_yaw'] or self.pretrain,
generators_count=src_generators_count ),
fs_aug = None
if ct_mode == 'fs-aug':
fs_aug = 'fs-aug'
SampleGeneratorFace(training_data_dst_path, debug=self.is_debug(), batch_size=self.get_batch_size(),
sample_process_options=SampleProcessor.Options(random_flip=self.random_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},
output_sample_types = [ {'sample_type': SampleProcessor.SampleType.FACE_IMAGE,'warp':random_warp, 'transform':True, 'channel_type' : SampleProcessor.ChannelType.BGR, 'ct_mode': fs_aug, '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': fs_aug, '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_EYES, 'face_type':self.face_type, 'data_format':nn.data_format, 'resolution': resolution},
],
uniform_yaw_distribution=self.options['uniform_yaw'] or self.pretrain,

11
samplelib/SampleProcessor.py
View file

@ -197,10 +197,13 @@ class SampleProcessor(object):
img = cv2.resize( img, (resolution, resolution), interpolation=cv2.INTER_CUBIC )
# Apply random color transfer
if ct_mode is not None and ct_sample is not None:
if ct_sample_bgr is None:
ct_sample_bgr = ct_sample.load_bgr()
img = imagelib.color_transfer (ct_mode, img, cv2.resize( ct_sample_bgr, (resolution,resolution), interpolation=cv2.INTER_LINEAR ) )
if ct_mode is not None and ct_sample is not None or ct_mode == 'fs-aug':
if 'fs-aug':
img = imagelib.color_augmentation(img)
elif:
if ct_sample_bgr is None:
ct_sample_bgr = ct_sample.load_bgr()
img = imagelib.color_transfer (ct_mode, img, cv2.resize( ct_sample_bgr, (resolution,resolution), interpolation=cv2.INTER_LINEAR ) )
img = imagelib.warp_by_params (params_per_resolution[resolution], img, warp, transform, can_flip=True, border_replicate=border_replicate)