def remove_xseg_labels(input_path):
if not input_path.exists():
raise ValueError(f'{input_path} not found. Please ensure it exists.')
io.log_info(f'Processing folder {input_path}')
io.log_info('!!! WARNING : LABELED XSEG POLYGONS WILL BE REMOVED FROM THE FRAMES !!!')
io.log_info('!!! WARNING : LABELED XSEG POLYGONS WILL BE REMOVED FROM THE FRAMES !!!')
io.log_info('!!! WARNING : LABELED XSEG POLYGONS WILL BE REMOVED FROM THE FRAMES !!!')
io.input_str('Press enter to continue.')
images_paths = pathex.get_image_paths(input_path, return_Path_class=True)
files_processed = 0
for filepath in io.progress_bar_generator(images_paths, "Processing"):
dflimg = DFLIMG.load(filepath)
if dflimg is None or not dflimg.has_data():
io.log_info(f'{filepath} is not a DFLIMG')
continue
if dflimg.has_seg_ie_polys():
dflimg.set_seg_ie_polys(None)
dflimg.save()
files_processed += 1
io.log_info(f'Files processed: {files_processed}')
def ask_choose_device_idxs(choose_only_one=False, allow_cpu=True, suggest_best_multi_gpu=False, suggest_all_gpu=False, return_device_config=False):
devices = Devices.getDevices()
if len(devices) == 0:
return []
all_devices_indexes = [device.index for device in devices]
if choose_only_one:
suggest_best_multi_gpu = False
suggest_all_gpu = False
if suggest_all_gpu:
best_device_indexes = all_devices_indexes
elif suggest_best_multi_gpu:
best_device_indexes = [device.index for device in devices.get_equal_devices(devices.get_best_device()) ]
else:
best_device_indexes = [ devices.get_best_device().index ]
best_device_indexes = ",".join([str(x) for x in best_device_indexes])
io.log_info ("")
if choose_only_one:
io.log_info ("Choose one GPU idx.")
else:
io.log_info ("Choose one or several GPU idxs (separated by comma).")
io.log_info ("")
if allow_cpu:
io.log_info ("[CPU] : CPU")
for device in devices:
io.log_info (f" [{device.index}] : {device.name}")
io.log_info ("")
while True:
try:
if choose_only_one:
choosed_idxs = io.input_str("Which GPU index to choose?", best_device_indexes)
else:
choosed_idxs = io.input_str("Which GPU indexes to choose?", best_device_indexes)
if allow_cpu and choosed_idxs.lower() == "cpu":
choosed_idxs = []
break
choosed_idxs = [ int(x) for x in choosed_idxs.split(',') ]
if choose_only_one:
if len(choosed_idxs) == 1:
break
else:
if all( [idx in all_devices_indexes for idx in choosed_idxs] ):
break
except:
pass
io.log_info ("")
if return_device_config:
return nn.DeviceConfig.GPUIndexes(choosed_idxs)
else:
return choosed_idxs
def denoise_image_sequence(input_dir, ext=None, factor=None):
input_path = Path(input_dir)
if not input_path.exists():
io.log_err("input_dir not found.")
return
if ext is None:
ext = io.input_str("Input image format (extension)", "png")
if factor is None:
factor = np.clip(io.input_int("Denoise factor?", 5, add_info="1-20"),
1, 20)
kwargs = {}
if ext == 'jpg':
kwargs.update({'q:v': '2'})
job = (ffmpeg.input(str(input_path / ('%5d.' + ext))).filter(
"hqdn3d", factor, factor, 5,
5).output(str(input_path / ('%5d.' + ext)), **kwargs))
try:
job = job.run()
except:
io.log_err("ffmpeg fail, job commandline:" + str(job.compile()))
def extract_video(input_file,
output_dir,
output_ext=None,
fps=None,
start_frame=None,
end_frame=None):
input_file_path = Path(input_file)
output_path = Path(output_dir)
if not output_path.exists():
output_path.mkdir(exist_ok=True)
if input_file_path.suffix == '.*':
input_file_path = pathex.get_first_file_by_stem(
input_file_path.parent, input_file_path.stem)
else:
if not input_file_path.exists():
input_file_path = None
if input_file_path is None:
io.log_err("input_file not found.")
return
if fps is None:
fps = io.input_int(
"Enter FPS",
0,
help_message=
"How many frames of every second of the video will be extracted. 0 - full fps"
)
if output_ext is None:
output_ext = io.input_str(
"Output image format",
"png", ["png", "jpg"],
help_message=
"png is lossless, but extraction is x10 slower for HDD, requires x10 more disk space than jpg."
)
for filename in pathex.get_image_paths(output_path, ['.' + output_ext]):
Path(filename).unlink()
job = ffmpeg.input(str(input_file_path))
if start_frame is not None and end_frame is not None:
job = job.trim(start_frame=start_frame, end_frame=end_frame)
kwargs = {'pix_fmt': 'rgb24'}
if fps != 0:
kwargs.update({'r': str(fps)})
if output_ext == 'jpg':
kwargs.update({'q:v': '2'}) #highest quality for jpg
job = job.output(str(output_path / ('%5d.' + output_ext)), **kwargs)
try:
job = job.run()
except:
io.log_err("ffmpeg fail, job commandline:" + str(job.compile()))
def cut_video(input_file,
from_time=None,
to_time=None,
audio_track_id=None,
bitrate=None):
input_file_path = Path(input_file)
if input_file_path is None:
io.log_err("input_file not found.")
return
output_file_path = input_file_path.parent / (
input_file_path.stem + "_cut" + input_file_path.suffix)
if from_time is None:
from_time = io.input_str("From time", "00:00:00.000")
if to_time is None:
to_time = io.input_str("To time", "00:00:00.000")
if audio_track_id is None:
audio_track_id = io.input_int("Specify audio track id.", 0)
if bitrate is None:
bitrate = max(1, io.input_int("输出码率 Bitrate of output file in MB/s",
25))
kwargs = {
"c:v": "libx264",
"b:v": "%dM" % (bitrate),
"pix_fmt": "yuv420p",
}
job = ffmpeg.input(str(input_file_path), ss=from_time, to=to_time)
job_v = job['v:0']
job_a = job['a:' + str(audio_track_id) + '?']
job = ffmpeg.output(job_v, job_a, str(output_file_path),
**kwargs).overwrite_output()
try:
job = job.run()
except:
io.log_err("ffmpeg fail, job commandline:" + str(job.compile()))
def ask_settings(self):
s = """Choose mode: \n"""
for key in mode_dict.keys():
s += f"""({key}) {mode_dict[key]}\n"""
io.log_info(s)
mode = io.input_int ("", mode_str_dict.get(self.default_mode, 1) )
self.mode = mode_dict.get (mode, self.default_mode )
if 'raw' not in self.mode:
if self.mode == 'hist-match':
self.masked_hist_match = io.input_bool("Masked hist match?", True)
if self.mode == 'hist-match' or self.mode == 'seamless-hist-match':
self.hist_match_threshold = np.clip ( io.input_int("Hist match threshold", 255, add_info="0..255"), 0, 255)
if self.face_type == FaceType.FULL:
s = """Choose mask mode: \n"""
for key in full_face_mask_mode_dict.keys():
s += f"""({key}) {full_face_mask_mode_dict[key]}\n"""
io.log_info(s)
self.mask_mode = io.input_int ("", 1, valid_list=full_face_mask_mode_dict.keys(), help_message="If you learned the mask, then option 1 should be choosed. 'dst' mask is raw shaky mask from dst aligned images. 'FAN-prd' - using super smooth mask by pretrained FAN-model from predicted face. 'FAN-dst' - using super smooth mask by pretrained FAN-model from dst face. 'FAN-prd*FAN-dst' or 'learned*FAN-prd*FAN-dst' - using multiplied masks.")
else:
s = """Choose mask mode: \n"""
for key in half_face_mask_mode_dict.keys():
s += f"""({key}) {half_face_mask_mode_dict[key]}\n"""
io.log_info(s)
self.mask_mode = io.input_int ("", 1, valid_list=half_face_mask_mode_dict.keys(), help_message="If you learned the mask, then option 1 should be choosed. 'dst' mask is raw shaky mask from dst aligned images.")
if 'raw' not in self.mode:
self.erode_mask_modifier = np.clip ( io.input_int ("Choose erode mask modifier", 0, add_info="-400..400"), -400, 400)
self.blur_mask_modifier = np.clip ( io.input_int ("Choose blur mask modifier", 0, add_info="0..400"), 0, 400)
self.motion_blur_power = np.clip ( io.input_int ("Choose motion blur power", 0, add_info="0..100"), 0, 100)
self.output_face_scale = np.clip (io.input_int ("Choose output face scale modifier", 0, add_info="-50..50" ), -50, 50)
if 'raw' not in self.mode:
self.color_transfer_mode = io.input_str ( "Color transfer to predicted face", None, valid_list=list(ctm_str_dict.keys())[1:] )
self.color_transfer_mode = ctm_str_dict[self.color_transfer_mode]
super().ask_settings()
self.super_resolution_power = np.clip ( io.input_int ("Choose super resolution power", 0, add_info="0..100", help_message="Enhance details by applying superresolution network."), 0, 100)
if 'raw' not in self.mode:
self.image_denoise_power = np.clip ( io.input_int ("Choose image degrade by denoise power", 0, add_info="0..500"), 0, 500)
self.bicubic_degrade_power = np.clip ( io.input_int ("Choose image degrade by bicubic rescale power", 0, add_info="0..100"), 0, 100)
self.color_degrade_power = np.clip ( io.input_int ("Degrade color power of final image", 0, add_info="0..100"), 0, 100)
io.log_info ("")
def on_initialize_options(self):
ask_override = self.ask_override()
if not self.is_first_run() and ask_override:
if io.input_bool(f"Restart training?", False, help_message="Reset model weights and start training from scratch."):
self.set_iter(0)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'wf')
if self.is_first_run():
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['h','mf','f','wf','head'], help_message="Half / mid face / full face / whole face / head. Choose the same as your deepfake model.").lower()
if self.is_first_run() or ask_override:
self.ask_batch_size(4, range=[2,16])
def process_folder ( dirpath):
image_size = io.input_int(f"New image size", 512, valid_range=[128,2048])
face_type = io.input_str ("Change face type", 'same', ['h','mf','f','wf','head','same']).lower()
if face_type == 'same':
face_type = None
else:
face_type = {'h' : FaceType.HALF,
'mf' : FaceType.MID_FULL,
'f' : FaceType.FULL,
'wf' : FaceType.WHOLE_FACE,
'head' : FaceType.HEAD}[face_type]
output_dirpath = dirpath.parent / (dirpath.name + '_resized')
output_dirpath.mkdir (exist_ok=True, parents=True)
dirpath_parts = '/'.join( dirpath.parts[-2:])
output_dirpath_parts = '/'.join( output_dirpath.parts[-2:] )
io.log_info (f"Resizing faceset in {dirpath_parts}")
io.log_info ( f"Processing to {output_dirpath_parts}")
output_images_paths = pathex.get_image_paths(output_dirpath)
if len(output_images_paths) > 0:
for filename in output_images_paths:
Path(filename).unlink()
image_paths = [Path(x) for x in pathex.get_image_paths( dirpath )]
result = FacesetResizerSubprocessor ( image_paths, output_dirpath, image_size, face_type).run()
is_merge = io.input_bool (f"\r\nMerge {output_dirpath_parts} to {dirpath_parts} ?", True)
if is_merge:
io.log_info (f"Copying processed files to {dirpath_parts}")
for (filepath, output_filepath) in result:
try:
shutil.copy (output_filepath, filepath)
except:
pass
io.log_info (f"Removing {output_dirpath_parts}")
shutil.rmtree(output_dirpath)
def ask_settings(self):
s = """Choose mode: \n"""
for key in mode_dict.keys():
s += f"""({key}) {mode_dict[key]}\n"""
io.log_info(s)
mode = io.input_int ("", mode_str_dict.get(self.default_mode, 1) )
self.mode = mode_dict.get (mode, self.default_mode )
if 'raw' not in self.mode:
if self.mode == 'hist-match':
self.masked_hist_match = io.input_bool("Masked hist match?", True)
if self.mode == 'hist-match' or self.mode == 'seamless-hist-match':
self.hist_match_threshold = np.clip ( io.input_int("Hist match threshold", 255, add_info="0..255"), 0, 255)
s = """Choose mask mode: \n"""
for key in mask_mode_dict.keys():
s += f"""({key}) {mask_mode_dict[key]}\n"""
io.log_info(s)
self.mask_mode = io.input_int ("", 4, valid_list=mask_mode_dict.keys() ) #
if 'raw' not in self.mode:
self.erode_mask_modifier = np.clip ( io.input_int ("Choose erode mask modifier", 0, add_info="-400..400"), -400, 400)
self.blur_mask_modifier = np.clip ( io.input_int ("Choose blur mask modifier", 0, add_info="0..400"), 0, 400)
self.motion_blur_power = np.clip ( io.input_int ("Choose motion blur power", 0, add_info="0..100"), 0, 100)
self.output_face_scale = np.clip (io.input_int ("Choose output face scale modifier", 0, add_info="-50..50" ), -50, 50)
if 'raw' not in self.mode:
self.color_transfer_mode = io.input_str ( "Color transfer to predicted face", "rct", valid_list=list(ctm_str_dict.keys())[1:] )
self.color_transfer_mode = ctm_str_dict[self.color_transfer_mode] #
super().ask_settings()
self.super_resolution_power = np.clip ( io.input_int ("Choose super resolution power", 0, add_info="0..100", help_message="Enhance details by applying superresolution network."), 0, 100)
if 'raw' not in self.mode:
self.image_denoise_power = np.clip ( io.input_int ("Choose image degrade by denoise power", 0, add_info="0..500"), 0, 500)
self.bicubic_degrade_power = np.clip ( io.input_int ("Choose image degrade by bicubic rescale power", 0, add_info="0..100"), 0, 100)
self.color_degrade_power = np.clip ( io.input_int ("Degrade color power of final image", 0, add_info="0..100"), 0, 100)
io.log_info ("")
def change_workspace():
wss = []
for f in os.listdir(get_root_path()):
fpath = os.path.join(get_root_path(), f)
if os.path.isfile(fpath) and f.startswith("@workspace"):
os.remove(fpath)
elif os.path.isdir(fpath) and f.startswith("workspace"):
wss.append(f)
inputs = "1234567890"[0:len(wss)]
for i in range(0, len(wss)):
io.log_info("[ %s ] %s" % (inputs[i], wss[i]))
no = io.input_str("Select Workspace:", 1)[0]
idx = inputs.find(no)
if idx < 0:
raise Exception("Invalid Idx " + no)
ws = wss[idx]
io.log_info("Select " + ws)
f = open(os.path.join(get_root_path(), "@" + ws), 'w')
f.write(ws)
f.close()
def on_initialize_options(self):
ask_override = self.ask_override()
if not self.is_first_run() and ask_override:
if io.input_bool(
f"Restart training?",
False,
help_message=
"Reset model weights and start training from scratch."):
self.set_iter(0)
default_face_type = self.options[
'face_type'] = self.load_or_def_option('face_type', 'wf')
default_pretrain = self.options['pretrain'] = self.load_or_def_option(
'pretrain', False)
if self.is_first_run():
self.options['face_type'] = io.input_str(
"Face type",
default_face_type, ['h', 'mf', 'f', 'wf', 'head'],
help_message=
"Half / mid face / full face / whole face / head. Choose the same as your deepfake model."
).lower()
if self.is_first_run() or ask_override:
self.ask_batch_size(4, range=[2, 16])
self.options['pretrain'] = io.input_bool("Enable pretraining mode",
default_pretrain)
if not self.is_exporting and (
self.options['pretrain']
and self.get_pretraining_data_path() is None):
raise Exception("pretraining_data_path is not defined")
self.pretrain_just_disabled = (default_pretrain == True
and self.options['pretrain'] == False)
def main(
detector=None,
input_path=None,
output_path=None,
output_debug=None,
manual_fix=False,
manual_output_debug_fix=False,
manual_window_size=1368,
face_type='full_face',
max_faces_from_image=None,
image_size=None,
jpeg_quality=None,
cpu_only=False,
force_gpu_idxs=None,
):
if not input_path.exists():
io.log_err('Input directory not found. Please ensure it exists.')
return
if not output_path.exists():
output_path.mkdir(parents=True, exist_ok=True)
if face_type is not None:
face_type = FaceType.fromString(face_type)
if face_type is None:
if manual_output_debug_fix:
files = pathex.get_image_paths(output_path)
if len(files) != 0:
dflimg = DFLIMG.load(Path(files[0]))
if dflimg is not None and dflimg.has_data():
face_type = FaceType.fromString(dflimg.get_face_type())
input_image_paths = pathex.get_image_unique_filestem_paths(
input_path, verbose_print_func=io.log_info)
output_images_paths = pathex.get_image_paths(output_path)
output_debug_path = output_path.parent / (output_path.name + '_debug')
continue_extraction = False
if not manual_output_debug_fix and len(output_images_paths) > 0:
if len(output_images_paths) > 128:
continue_extraction = io.input_bool(
"Continue extraction?",
True,
help_message=
"Extraction can be continued, but you must specify the same options again."
)
if len(output_images_paths) > 128 and continue_extraction:
try:
input_image_paths = input_image_paths[
[Path(x).stem for x in input_image_paths].
index(Path(output_images_paths[-128]).stem.split('_')[0]):]
except:
io.log_err(
"Error in fetching the last index. Extraction cannot be continued."
)
return
elif input_path != output_path:
io.input(
f"\n WARNING !!! \n {output_path} contains files! \n They will be deleted. \n Press enter to continue.\n"
)
for filename in output_images_paths:
Path(filename).unlink()
device_config = nn.DeviceConfig.GPUIndexes( force_gpu_idxs or nn.ask_choose_device_idxs(choose_only_one=detector=='manual', suggest_all_gpu=True) ) \
if not cpu_only else nn.DeviceConfig.CPU()
if face_type is None:
face_type = io.input_str(
"Face type",
'wf', ['f', 'wf', 'head'],
help_message=
"Full face / whole face / head. 'Whole face' covers full area of face include forehead. 'head' covers full head, but requires XSeg for src and dst faceset."
).lower()
face_type = {
'f': FaceType.FULL,
'wf': FaceType.WHOLE_FACE,
'head': FaceType.HEAD
}[face_type]
if max_faces_from_image is None:
max_faces_from_image = io.input_int(
f"Max number of faces from image",
0,
help_message=
"If you extract a src faceset that has frames with a large number of faces, it is advisable to set max faces to 3 to speed up extraction. 0 - unlimited"
)
if image_size is None:
image_size = io.input_int(
f"Image size",
512 if face_type < FaceType.HEAD else 768,
valid_range=[256, 2048],
help_message=
"Output image size. The higher image size, the worse face-enhancer works. Use higher than 512 value only if the source image is sharp enough and the face does not need to be enhanced."
)
if jpeg_quality is None:
jpeg_quality = io.input_int(
f"Jpeg quality",
90,
valid_range=[1, 100],
help_message=
"Jpeg quality. The higher jpeg quality the larger the output file size."
)
if detector is None:
io.log_info("Choose detector type.")
io.log_info("[0] S3FD")
io.log_info("[1] manual")
detector = {0: 's3fd', 1: 'manual'}[io.input_int("", 0, [0, 1])]
if output_debug is None:
output_debug = io.input_bool(
f"Write debug images to {output_debug_path.name}?", False)
if output_debug:
output_debug_path.mkdir(parents=True, exist_ok=True)
if manual_output_debug_fix:
if not output_debug_path.exists():
io.log_err(
f'{output_debug_path} not found. Re-extract faces with "Write debug images" option.'
)
return
else:
detector = 'manual'
io.log_info(
'Performing re-extract frames which were deleted from _debug directory.'
)
input_image_paths = DeletedFilesSearcherSubprocessor(
input_image_paths,
pathex.get_image_paths(output_debug_path)).run()
input_image_paths = sorted(input_image_paths)
io.log_info('Found %d images.' % (len(input_image_paths)))
else:
if not continue_extraction and output_debug_path.exists():
for filename in pathex.get_image_paths(output_debug_path):
Path(filename).unlink()
images_found = len(input_image_paths)
faces_detected = 0
if images_found != 0:
if detector == 'manual':
io.log_info('Performing manual extract...')
data = ExtractSubprocessor(
[
ExtractSubprocessor.Data(Path(filename))
for filename in input_image_paths
],
'landmarks-manual',
image_size,
jpeg_quality,
face_type,
output_debug_path if output_debug else None,
manual_window_size=manual_window_size,
device_config=device_config).run()
io.log_info('Performing 3rd pass...')
data = ExtractSubprocessor(
data,
'final',
image_size,
jpeg_quality,
face_type,
output_debug_path if output_debug else None,
final_output_path=output_path,
device_config=device_config).run()
else:
io.log_info('Extracting faces...')
data = ExtractSubprocessor(
[
ExtractSubprocessor.Data(Path(filename))
for filename in input_image_paths
],
'all',
image_size,
jpeg_quality,
face_type,
output_debug_path if output_debug else None,
max_faces_from_image=max_faces_from_image,
final_output_path=output_path,
device_config=device_config).run()
faces_detected += sum([d.faces_detected for d in data])
if manual_fix:
if all(np.array([d.faces_detected > 0 for d in data]) == True):
io.log_info('All faces are detected, manual fix not needed.')
else:
fix_data = [
ExtractSubprocessor.Data(d.filepath) for d in data
if d.faces_detected == 0
]
io.log_info('Performing manual fix for %d images...' %
(len(fix_data)))
fix_data = ExtractSubprocessor(
fix_data,
'landmarks-manual',
image_size,
jpeg_quality,
face_type,
output_debug_path if output_debug else None,
manual_window_size=manual_window_size,
device_config=device_config).run()
fix_data = ExtractSubprocessor(
fix_data,
'final',
image_size,
jpeg_quality,
face_type,
output_debug_path if output_debug else None,
final_output_path=output_path,
device_config=device_config).run()
faces_detected += sum([d.faces_detected for d in fix_data])
io.log_info('-------------------------')
io.log_info('Images found: %d' % (images_found))
io.log_info('Faces detected: %d' % (faces_detected))
io.log_info('-------------------------')
def on_initialize_options(self):
default_resolution = self.options['resolution'] = self.load_or_def_option('resolution', 224)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'wf')
default_models_opt_on_gpu = self.options['models_opt_on_gpu'] = self.load_or_def_option('models_opt_on_gpu', True)
default_ae_dims = self.options['ae_dims'] = self.load_or_def_option('ae_dims', 256)
default_inter_dims = self.options['inter_dims'] = self.load_or_def_option('inter_dims', 1024)
default_e_dims = self.options['e_dims'] = self.load_or_def_option('e_dims', 64)
default_d_dims = self.options['d_dims'] = self.options.get('d_dims', None)
default_d_mask_dims = self.options['d_mask_dims'] = self.options.get('d_mask_dims', None)
default_morph_factor = self.options['morph_factor'] = self.options.get('morph_factor', 0.5)
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_lr_dropout = self.options['lr_dropout'] = self.load_or_def_option('lr_dropout', 'n')
default_random_warp = self.options['random_warp'] = self.load_or_def_option('random_warp', True)
default_ct_mode = self.options['ct_mode'] = self.load_or_def_option('ct_mode', 'none')
default_clipgrad = self.options['clipgrad'] = self.load_or_def_option('clipgrad', False)
ask_override = self.ask_override()
if self.is_first_run() or ask_override:
self.ask_autobackup_hour()
self.ask_write_preview_history()
self.ask_target_iter()
self.ask_random_src_flip()
self.ask_random_dst_flip()
self.ask_batch_size(8)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-640", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 32 .")
resolution = np.clip ( (resolution // 32) * 32, 64, 640)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['f','wf','head'], help_message="whole face / head").lower()
default_d_dims = self.options['d_dims'] = self.load_or_def_option('d_dims', 64)
default_d_mask_dims = default_d_dims // 3
default_d_mask_dims += default_d_mask_dims % 2
default_d_mask_dims = self.options['d_mask_dims'] = self.load_or_def_option('d_mask_dims', default_d_mask_dims)
if self.is_first_run():
self.options['ae_dims'] = np.clip ( io.input_int("AutoEncoder dimensions", default_ae_dims, add_info="32-1024", help_message="All face information will packed to AE dims. If amount of AE dims are not enough, then for example closed eyes will not be recognized. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 1024 )
self.options['inter_dims'] = np.clip ( io.input_int("Inter dimensions", default_inter_dims, add_info="32-2048", help_message="Should be equal or more than AutoEncoder dimensions. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 2048 )
e_dims = np.clip ( io.input_int("Encoder dimensions", default_e_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['e_dims'] = e_dims + e_dims % 2
d_dims = np.clip ( io.input_int("Decoder dimensions", default_d_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['d_dims'] = d_dims + d_dims % 2
d_mask_dims = np.clip ( io.input_int("Decoder mask dimensions", default_d_mask_dims, add_info="16-256", help_message="Typical mask dimensions = decoder dimensions / 3. If you manually cut out obstacles from the dst mask, you can increase this parameter to achieve better quality." ), 16, 256 )
self.options['d_mask_dims'] = d_mask_dims + d_mask_dims % 2
morph_factor = np.clip ( io.input_number ("Morph factor.", default_morph_factor, add_info="0.1 .. 0.5", help_message="Typical fine value is 0.5"), 0.1, 0.5 )
self.options['morph_factor'] = morph_factor
if self.is_first_run() or ask_override:
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['lr_dropout'] = io.input_str (f"Use learning rate dropout", default_lr_dropout, ['n','y','cpu'], help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations. Enabled it before `disable random warp` and before GAN. \nn - disabled.\ny - enabled\ncpu - enabled on CPU. This allows not to use extra VRAM, sacrificing 20% time of iteration.")
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)
default_gan_dims = self.options['gan_dims'] = self.load_or_def_option('gan_dims', 16)
if self.is_first_run() or ask_override:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['gan_power'] = np.clip ( io.input_number ("GAN power", default_gan_power, add_info="0.0 .. 5.0", help_message="Forces the neural network to learn small details of the face. Enable it only when the face is trained enough with random_warp(off), and don't disable. The higher the value, the higher the chances of artifacts. Typical fine value is 0.1"), 0.0, 5.0 )
if self.options['gan_power'] != 0.0:
gan_patch_size = np.clip ( io.input_int("GAN patch size", default_gan_patch_size, add_info="3-640", help_message="The higher patch size, the higher the quality, the more VRAM is required. You can get sharper edges even at the lowest setting. Typical fine value is resolution / 8." ), 3, 640 )
self.options['gan_patch_size'] = gan_patch_size
gan_dims = np.clip ( io.input_int("GAN dimensions", default_gan_dims, add_info="4-512", help_message="The dimensions of the GAN network. The higher dimensions, the more VRAM is required. You can get sharper edges even at the lowest setting. Typical fine value is 16." ), 4, 512 )
self.options['gan_dims'] = gan_dims
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. If src faceset is deverse enough, then lct mode is fine in most cases.")
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.gan_model_changed = (default_gan_patch_size != self.options['gan_patch_size']) or (default_gan_dims != self.options['gan_dims'])
def apply_xseg(input_path, model_path):
if not input_path.exists():
raise ValueError(f'{input_path} not found. Please ensure it exists.')
if not model_path.exists():
raise ValueError(f'{model_path} not found. Please ensure it exists.')
face_type = None
model_dat = model_path / 'XSeg_data.dat'
if model_dat.exists():
dat = pickle.loads( model_dat.read_bytes() )
dat_options = dat.get('options', None)
if dat_options is not None:
face_type = dat_options.get('face_type', None)
if face_type is None:
face_type = io.input_str ("XSeg model face type", 'same', ['h','mf','f','wf','head','same'], help_message="Specify face type of trained XSeg model. For example if XSeg model trained as WF, but faceset is HEAD, specify WF to apply xseg only on WF part of HEAD. Default is 'same'").lower()
if face_type == 'same':
face_type = None
if face_type is not None:
face_type = {'h' : FaceType.HALF,
'mf' : FaceType.MID_FULL,
'f' : FaceType.FULL,
'wf' : FaceType.WHOLE_FACE,
'head' : FaceType.HEAD}[face_type]
io.log_info(f'Applying trained XSeg model to {input_path.name}/ folder.')
device_config = nn.DeviceConfig.ask_choose_device(choose_only_one=True)
nn.initialize(device_config)
xseg = XSegNet(name='XSeg',
load_weights=True,
weights_file_root=model_path,
data_format=nn.data_format,
raise_on_no_model_files=True)
xseg_res = xseg.get_resolution()
images_paths = pathex.get_image_paths(input_path, return_Path_class=True)
for filepath in io.progress_bar_generator(images_paths, "Processing"):
dflimg = DFLIMG.load(filepath)
if dflimg is None or not dflimg.has_data():
io.log_info(f'{filepath} is not a DFLIMG')
continue
img = cv2_imread(filepath).astype(np.float32) / 255.0
h,w,c = img.shape
img_face_type = FaceType.fromString( dflimg.get_face_type() )
if face_type is not None and img_face_type != face_type:
lmrks = dflimg.get_source_landmarks()
fmat = LandmarksProcessor.get_transform_mat(lmrks, w, face_type)
imat = LandmarksProcessor.get_transform_mat(lmrks, w, img_face_type)
g_p = LandmarksProcessor.transform_points (np.float32([(0,0),(w,0),(0,w) ]), fmat, True)
g_p2 = LandmarksProcessor.transform_points (g_p, imat)
mat = cv2.getAffineTransform( g_p2, np.float32([(0,0),(w,0),(0,w) ]) )
img = cv2.warpAffine(img, mat, (w, w), cv2.INTER_LANCZOS4)
img = cv2.resize(img, (xseg_res, xseg_res), interpolation=cv2.INTER_LANCZOS4)
else:
if w != xseg_res:
img = cv2.resize( img, (xseg_res,xseg_res), interpolation=cv2.INTER_LANCZOS4 )
if len(img.shape) == 2:
img = img[...,None]
mask = xseg.extract(img)
if face_type is not None and img_face_type != face_type:
mask = cv2.resize(mask, (w, w), interpolation=cv2.INTER_LANCZOS4)
mask = cv2.warpAffine( mask, mat, (w,w), np.zeros( (h,w,c), dtype=np.float), cv2.WARP_INVERSE_MAP | cv2.INTER_LANCZOS4)
mask = cv2.resize(mask, (xseg_res, xseg_res), interpolation=cv2.INTER_LANCZOS4)
mask[mask < 0.5]=0
mask[mask >= 0.5]=1
dflimg.set_xseg_mask(mask)
dflimg.save()
def ask_session_name(self, default_value=""):
default_session_name = self.options['session_name'] = self.load_or_def_option('session_name', default_value)
self.options['session_name'] = io.input_str("Session name", default_session_name, help_message="String to refer back to in summary.txt and in autobackup foldername")
def extract_video(input_file, output_dir, output_ext=None, fps=None):
input_file_path = Path(input_file)
output_path = Path(output_dir)
if not output_path.exists():
output_path.mkdir(exist_ok=True)
InfoNotifier.InfoNotifier.g_progress_info.append(
"\n视频帧输出目录: " + str(Path(output_path).absolute()))
if input_file_path.suffix == '.*':
input_file_path = pathex.get_first_file_by_stem(
input_file_path.parent, input_file_path.stem)
else:
if not input_file_path.exists():
input_file_path = None
InfoNotifier.InfoNotifier.g_progress_info.append("\n视频输入路径:" +
str(input_file_path))
if input_file_path is None:
io.log_err("input_file not found.")
InfoNotifier.InfoNotifier.g_progress_info.append("\n视频输入路径不存在")
return
if fps is None:
fps = io.input_int(
"Enter FPS",
0,
help_message=
"How many frames of every second of the video will be extracted. 0 - full fps"
)
InfoNotifier.InfoNotifier.g_progress_info.append("\n视频帧抽取频率: full fps")
if output_ext is None:
output_ext = io.input_str(
"Output image format",
"png", ["png", "jpg"],
help_message=
"png is lossless, but extraction is x10 slower for HDD, requires x10 more disk space than jpg."
)
InfoNotifier.InfoNotifier.g_progress_info.append("\n视频帧输出格式频率: " +
output_ext)
filenames = pathex.get_image_paths(output_path, ['.' + output_ext])
if len(filenames) != 0:
InfoNotifier.InfoNotifier.g_progress_info.append(
"\n视频帧输出目录不为空, 该目录将被清空!")
for filename in filenames:
Path(filename).unlink()
QApplication.processEvents()
job = ffmpeg.input(str(input_file_path))
kwargs = {'pix_fmt': 'rgb24'}
if fps != 0:
kwargs.update({'r': str(fps)})
if output_ext == 'jpg':
kwargs.update({'q:v': '2'}) #highest quality for jpg
job = job.output(str(output_path / ('%5d.' + output_ext)), **kwargs)
try:
job, err = job.run(cmd=UIParamReflect.GlobalConfig.ffmpeg_cmd_path)
except:
io.log_err("ffmpeg fail, job commandline:" + str(job.compile()))
def main(
detector=None,
input_path=None,
output_path=None,
output_debug=None,
manual_fix=False,
manual_output_debug_fix=False,
manual_window_size=1368,
face_type='full_face',
max_faces_from_image=0,
cpu_only=False,
force_gpu_idxs=None,
):
if not input_path.exists():
io.log_err('Input directory not found. Please ensure it exists.')
return
if face_type is not None:
face_type = FaceType.fromString(face_type)
if face_type is None:
if manual_output_debug_fix and output_path.exists():
files = pathex.get_image_paths(output_path)
if len(files) != 0:
dflimg = DFLIMG.load(Path(files[0]))
if dflimg is not None and dflimg.has_data():
face_type = FaceType.fromString(dflimg.get_face_type())
if face_type is None:
face_type = io.input_str(
"Face type",
'wf', ['f', 'wf', 'head'],
help_message=
"Full face / whole face / head. 'Whole face' covers full area of face include forehead. 'head' covers full head, but requires XSeg for src and dst faceset."
).lower()
face_type = {
'f': FaceType.FULL,
'wf': FaceType.WHOLE_FACE,
'head': FaceType.HEAD
}[face_type]
image_size = 512 if face_type < FaceType.HEAD else 768
if detector is None:
io.log_info("Choose detector type.")
io.log_info("[0] S3FD")
io.log_info("[1] manual")
detector = {0: 's3fd', 1: 'manual'}[io.input_int("", 0, [0, 1])]
device_config = nn.DeviceConfig.GPUIndexes( force_gpu_idxs or nn.ask_choose_device_idxs(choose_only_one=detector=='manual', suggest_all_gpu=True) ) \
if not cpu_only else nn.DeviceConfig.CPU()
output_debug_path = output_path.parent / (output_path.name + '_debug')
if output_debug is None:
output_debug = io.input_bool(
f"Write debug images to {output_debug_path.name}?", False)
if output_path.exists():
if not manual_output_debug_fix and input_path != output_path:
output_images_paths = pathex.get_image_paths(output_path)
if len(output_images_paths) > 0:
io.input(
f"\n WARNING !!! \n {output_path} contains files! \n They will be deleted. \n Press enter to continue.\n"
)
for filename in output_images_paths:
Path(filename).unlink()
else:
output_path.mkdir(parents=True, exist_ok=True)
input_path_image_paths = pathex.get_image_unique_filestem_paths(
input_path, verbose_print_func=io.log_info)
if manual_output_debug_fix:
if not output_debug_path.exists():
io.log_err(
f'{output_debug_path} not found. Re-extract faces with "Write debug images" option.'
)
return
else:
detector = 'manual'
io.log_info(
'Performing re-extract frames which were deleted from _debug directory.'
)
input_path_image_paths = DeletedFilesSearcherSubprocessor(
input_path_image_paths,
pathex.get_image_paths(output_debug_path)).run()
input_path_image_paths = sorted(input_path_image_paths)
io.log_info('Found %d images.' % (len(input_path_image_paths)))
else:
if output_debug_path.exists():
for filename in pathex.get_image_paths(output_debug_path):
Path(filename).unlink()
else:
output_debug_path.mkdir(parents=True, exist_ok=True)
images_found = len(input_path_image_paths)
faces_detected = 0
if images_found != 0:
if detector == 'manual':
io.log_info('Performing manual extract...')
data = ExtractSubprocessor(
[
ExtractSubprocessor.Data(Path(filename))
for filename in input_path_image_paths
],
'landmarks-manual',
image_size,
face_type,
output_debug_path if output_debug else None,
manual_window_size=manual_window_size,
device_config=device_config).run()
io.log_info('Performing 3rd pass...')
data = ExtractSubprocessor(
data,
'final',
image_size,
face_type,
output_debug_path if output_debug else None,
final_output_path=output_path,
device_config=device_config).run()
else:
io.log_info('Extracting faces...')
data = ExtractSubprocessor(
[
ExtractSubprocessor.Data(Path(filename))
for filename in input_path_image_paths
],
'all',
image_size,
face_type,
output_debug_path if output_debug else None,
max_faces_from_image=max_faces_from_image,
final_output_path=output_path,
device_config=device_config).run()
faces_detected += sum([d.faces_detected for d in data])
if manual_fix:
if all(np.array([d.faces_detected > 0 for d in data]) == True):
io.log_info('All faces are detected, manual fix not needed.')
else:
fix_data = [
ExtractSubprocessor.Data(d.filepath) for d in data
if d.faces_detected == 0
]
io.log_info('Performing manual fix for %d images...' %
(len(fix_data)))
fix_data = ExtractSubprocessor(
fix_data,
'landmarks-manual',
image_size,
face_type,
output_debug_path if output_debug else None,
manual_window_size=manual_window_size,
device_config=device_config).run()
fix_data = ExtractSubprocessor(
fix_data,
'final',
image_size,
face_type,
output_debug_path if output_debug else None,
final_output_path=output_path,
device_config=device_config).run()
faces_detected += sum([d.faces_detected for d in fix_data])
io.log_info('-------------------------')
io.log_info('Images found: %d' % (images_found))
io.log_info('Faces detected: %d' % (faces_detected))
io.log_info('-------------------------')
def on_initialize_options(self):
device_config = nn.getCurrentDeviceConfig()
lowest_vram = 2
if len(device_config.devices) != 0:
lowest_vram = device_config.devices.get_worst_device().total_mem_gb
if lowest_vram >= 4:
suggest_batch_size = 8
else:
suggest_batch_size = 4
yn_str = {True:'y',False:'n'}
min_res = 64
max_res = 640
default_resolution = self.options['resolution'] = self.load_or_def_option('resolution', 224)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'wf')
default_models_opt_on_gpu = self.options['models_opt_on_gpu'] = self.load_or_def_option('models_opt_on_gpu', True)
default_ae_dims = self.options['ae_dims'] = self.load_or_def_option('ae_dims', 256)
default_e_dims = self.options['e_dims'] = self.load_or_def_option('e_dims', 64)
default_d_dims = self.options['d_dims'] = self.options.get('d_dims', None)
default_d_mask_dims = self.options['d_mask_dims'] = self.options.get('d_mask_dims', None)
default_morph_factor = self.options['morph_factor'] = self.options.get('morph_factor', 0.33)
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', True)
default_uniform_yaw = self.options['uniform_yaw'] = self.load_or_def_option('uniform_yaw', False)
lr_dropout = self.load_or_def_option('lr_dropout', 'n')
lr_dropout = {True:'y', False:'n'}.get(lr_dropout, lr_dropout) #backward comp
default_lr_dropout = self.options['lr_dropout'] = lr_dropout
default_random_warp = self.options['random_warp'] = self.load_or_def_option('random_warp', True)
default_ct_mode = self.options['ct_mode'] = self.load_or_def_option('ct_mode', 'none')
default_clipgrad = self.options['clipgrad'] = self.load_or_def_option('clipgrad', False)
default_pretrain = self.options['pretrain'] = self.load_or_def_option('pretrain', False)
ask_override = self.ask_override()
if self.is_first_run() or ask_override:
self.ask_autobackup_hour()
self.ask_write_preview_history()
self.ask_target_iter()
self.ask_random_src_flip()
self.ask_random_dst_flip()
self.ask_batch_size(suggest_batch_size)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-640", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 32 .")
resolution = np.clip ( (resolution // 32) * 32, min_res, max_res)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['wf','head'], help_message="whole face / head").lower()
default_d_dims = self.options['d_dims'] = self.load_or_def_option('d_dims', 64)
default_d_mask_dims = default_d_dims // 3
default_d_mask_dims += default_d_mask_dims % 2
default_d_mask_dims = self.options['d_mask_dims'] = self.load_or_def_option('d_mask_dims', default_d_mask_dims)
if self.is_first_run():
self.options['ae_dims'] = np.clip ( io.input_int("AutoEncoder dimensions", default_ae_dims, add_info="32-1024", help_message="All face information will packed to AE dims. If amount of AE dims are not enough, then for example closed eyes will not be recognized. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 1024 )
e_dims = np.clip ( io.input_int("Encoder dimensions", default_e_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['e_dims'] = e_dims + e_dims % 2
d_dims = np.clip ( io.input_int("Decoder dimensions", default_d_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['d_dims'] = d_dims + d_dims % 2
d_mask_dims = np.clip ( io.input_int("Decoder mask dimensions", default_d_mask_dims, add_info="16-256", help_message="Typical mask dimensions = decoder dimensions / 3. If you manually cut out obstacles from the dst mask, you can increase this parameter to achieve better quality." ), 16, 256 )
self.options['d_mask_dims'] = d_mask_dims + d_mask_dims % 2
morph_factor = np.clip ( io.input_number ("Morph factor.", default_morph_factor, add_info="0.1 .. 0.5", help_message="The smaller the value, the more src-like facial expressions will appear. The larger the value, the less space there is to train a large dst faceset in the neural network. Typical fine value is 0.33"), 0.1, 0.5 )
self.options['morph_factor'] = morph_factor
if self.is_first_run() or ask_override:
if self.options['face_type'] == 'wf' or self.options['face_type'] == 'head':
self.options['masked_training'] = io.input_bool ("Masked training", default_masked_training, help_message="This option is available only for 'whole_face' or 'head' type. Masked training clips training area to full_face mask or XSeg mask, thus network will train the faces properly.")
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.')
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)
default_gan_dims = self.options['gan_dims'] = self.load_or_def_option('gan_dims', 16)
if self.is_first_run() or ask_override:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
self.options['lr_dropout'] = io.input_str (f"Use learning rate dropout", default_lr_dropout, ['n','y','cpu'], help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations. Enabled it before `disable random warp` and before GAN. \nn - disabled.\ny - enabled\ncpu - enabled on CPU. This allows not to use extra VRAM, sacrificing 20% time of iteration.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['gan_power'] = np.clip ( io.input_number ("GAN power", default_gan_power, add_info="0.0 .. 1.0", help_message="Forces the neural network to learn small details of the face. Enable it only when the face is trained enough with lr_dropout(on) and random_warp(off), and don't disable. The higher the value, the higher the chances of artifacts. Typical fine value is 0.1"), 0.0, 1.0 )
if self.options['gan_power'] != 0.0:
gan_patch_size = np.clip ( io.input_int("GAN patch size", default_gan_patch_size, add_info="3-640", help_message="The higher patch size, the higher the quality, the more VRAM is required. You can get sharper edges even at the lowest setting. Typical fine value is resolution / 8." ), 3, 640 )
self.options['gan_patch_size'] = gan_patch_size
gan_dims = np.clip ( io.input_int("GAN dimensions", default_gan_dims, add_info="4-64", help_message="The dimensions of the GAN network. The higher dimensions, the more VRAM is required. You can get sharper edges even at the lowest setting. Typical fine value is 16." ), 4, 64 )
self.options['gan_dims'] = gan_dims
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['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. Forces random_warp=N, random_flips=Y, gan_power=0.0, lr_dropout=N, uniform_yaw=Y")
self.gan_model_changed = (default_gan_patch_size != self.options['gan_patch_size']) or (default_gan_dims != self.options['gan_dims'])
self.pretrain_just_disabled = (default_pretrain == True and self.options['pretrain'] == False)
def __init__(self, is_training=False,
saved_models_path=None,
training_data_src_path=None,
training_data_dst_path=None,
pretraining_data_path=None,
pretrained_model_path=None,
no_preview=False,
force_model_name=None,
force_gpu_idxs=None,
cpu_only=False,
debug=False,
force_model_class_name=None,
**kwargs):
self.is_training = is_training
self.saved_models_path = saved_models_path
self.training_data_src_path = training_data_src_path
self.training_data_dst_path = training_data_dst_path
self.pretraining_data_path = pretraining_data_path
self.pretrained_model_path = pretrained_model_path
self.no_preview = no_preview
self.debug = debug
self.model_class_name = model_class_name = Path(inspect.getmodule(self).__file__).parent.name.rsplit("_", 1)[1]
if force_model_class_name is None:
if force_model_name is not None:
self.model_name = force_model_name
else:
while True:
# gather all model dat files
saved_models_names = []
for filepath in pathex.get_file_paths(saved_models_path):
filepath_name = filepath.name
if filepath_name.endswith(f'{model_class_name}_data.dat'):
saved_models_names += [ (filepath_name.split('_')[0], os.path.getmtime(filepath)) ]
# sort by modified datetime
saved_models_names = sorted(saved_models_names, key=operator.itemgetter(1), reverse=True )
saved_models_names = [ x[0] for x in saved_models_names ]
if len(saved_models_names) != 0:
io.log_info ("Choose one of saved models, or enter a name to create a new model.")
io.log_info ("[r] : rename")
io.log_info ("[d] : delete")
io.log_info ("")
for i, model_name in enumerate(saved_models_names):
s = f"[{i}] : {model_name} "
if i == 0:
s += "- latest"
io.log_info (s)
inp = io.input_str(f"", "0", show_default_value=False )
model_idx = -1
try:
model_idx = np.clip ( int(inp), 0, len(saved_models_names)-1 )
except:
pass
if model_idx == -1:
if len(inp) == 1:
is_rename = inp[0] == 'r'
is_delete = inp[0] == 'd'
if is_rename or is_delete:
if len(saved_models_names) != 0:
if is_rename:
name = io.input_str(f"Enter the name of the model you want to rename")
elif is_delete:
name = io.input_str(f"Enter the name of the model you want to delete")
if name in saved_models_names:
if is_rename:
new_model_name = io.input_str(f"Enter new name of the model")
for filepath in pathex.get_paths(saved_models_path):
filepath_name = filepath.name
model_filename, remain_filename = filepath_name.split('_', 1)
if model_filename == name:
if is_rename:
new_filepath = filepath.parent / ( new_model_name + '_' + remain_filename )
filepath.rename (new_filepath)
elif is_delete:
filepath.unlink()
continue
self.model_name = inp
else:
self.model_name = saved_models_names[model_idx]
else:
self.model_name = io.input_str(f"No saved models found. Enter a name of a new model", "new")
self.model_name = self.model_name.replace('_', ' ')
break
self.model_name = self.model_name + '_' + self.model_class_name
else:
self.model_name = force_model_class_name
self.iter = 0
self.options = {}
self.loss_history = []
self.sample_for_preview = None
self.choosed_gpu_indexes = None
model_data = {}
self.model_data_path = Path( self.get_strpath_storage_for_file('data.dat') )
if self.model_data_path.exists():
io.log_info (f"Loading {self.model_name} model...")
model_data = pickle.loads ( self.model_data_path.read_bytes() )
self.iter = model_data.get('iter',0)
if self.iter != 0:
self.options = model_data['options']
self.loss_history = model_data.get('loss_history', [])
self.sample_for_preview = model_data.get('sample_for_preview', None)
self.choosed_gpu_indexes = model_data.get('choosed_gpu_indexes', None)
if self.is_first_run():
io.log_info ("\nModel first run.")
self.device_config = nn.DeviceConfig.GPUIndexes( force_gpu_idxs or nn.ask_choose_device_idxs(suggest_best_multi_gpu=True)) \
if not cpu_only else nn.DeviceConfig.CPU()
nn.initialize(self.device_config)
####
self.default_options_path = saved_models_path / f'{self.model_class_name}_default_options.dat'
self.default_options = {}
if self.default_options_path.exists():
try:
self.default_options = pickle.loads ( self.default_options_path.read_bytes() )
except:
pass
self.choose_preview_history = False
self.batch_size = self.load_or_def_option('batch_size', 1)
#####
io.input_skip_pending()
self.on_initialize_options()
if self.is_first_run():
# save as default options only for first run model initialize
self.default_options_path.write_bytes( pickle.dumps (self.options) )
self.autobackup_hour = self.options.get('autobackup_hour', 0)
self.write_preview_history = self.options.get('write_preview_history', False)
self.target_iter = self.options.get('target_iter',0)
self.random_flip = self.options.get('random_flip',True)
self.on_initialize()
self.options['batch_size'] = self.batch_size
if self.is_training:
self.preview_history_path = self.saved_models_path / ( f'{self.get_model_name()}_history' )
self.autobackups_path = self.saved_models_path / ( f'{self.get_model_name()}_autobackups' )
if self.write_preview_history or io.is_colab():
if not self.preview_history_path.exists():
self.preview_history_path.mkdir(exist_ok=True)
else:
if self.iter == 0:
for filename in pathex.get_image_paths(self.preview_history_path):
Path(filename).unlink()
if self.generator_list is None:
raise ValueError( 'You didnt set_training_data_generators()')
else:
for i, generator in enumerate(self.generator_list):
if not isinstance(generator, SampleGeneratorBase):
raise ValueError('training data generator is not subclass of SampleGeneratorBase')
self.update_sample_for_preview(choose_preview_history=self.choose_preview_history)
if self.autobackup_hour != 0:
self.autobackup_start_time = time.time()
if not self.autobackups_path.exists():
self.autobackups_path.mkdir(exist_ok=True)
io.log_info( self.get_summary_text() )
def on_initialize_options(self):
device_config = nn.getCurrentDeviceConfig()
lowest_vram = 2
if len(device_config.devices) != 0:
lowest_vram = device_config.devices.get_worst_device().total_mem_gb
if lowest_vram >= 4:
suggest_batch_size = 8
else:
suggest_batch_size = 4
yn_str = {True:'y',False:'n'}
ask_override = self.ask_override()
if self.is_first_run() or ask_override:
self.ask_autobackup_hour()
self.ask_write_preview_history()
self.ask_target_iter()
self.ask_random_flip()
self.ask_batch_size(suggest_batch_size)
default_resolution = self.options['resolution'] = self.load_or_def_option('resolution', 128)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'f')
default_models_opt_on_gpu = self.options['models_opt_on_gpu'] = self.load_or_def_option('models_opt_on_gpu', True)
default_archi = self.options['archi'] = self.load_or_def_option('archi', 'dfhd')
default_ae_dims = self.options['ae_dims'] = self.load_or_def_option('ae_dims', 256)
default_e_dims = self.options['e_dims'] = self.load_or_def_option('e_dims', 64)
default_d_dims = self.options['d_dims'] = self.load_or_def_option('d_dims', 64)
default_d_mask_dims = default_d_dims // 3
default_d_mask_dims += default_d_mask_dims % 2
default_d_mask_dims = self.options['d_mask_dims'] = self.load_or_def_option('d_mask_dims', default_d_mask_dims)
default_learn_mask = self.options['learn_mask'] = self.load_or_def_option('learn_mask', True)
default_lr_dropout = self.options['lr_dropout'] = self.load_or_def_option('lr_dropout', False)
default_random_warp = self.options['random_warp'] = self.load_or_def_option('random_warp', True)
default_true_face_training = self.options['true_face_training'] = self.load_or_def_option('true_face_training', False)
default_face_style_power = self.options['face_style_power'] = self.load_or_def_option('face_style_power', 0.0)
default_bg_style_power = self.options['bg_style_power'] = self.load_or_def_option('bg_style_power', 0.0)
default_ct_mode = self.options['ct_mode'] = self.load_or_def_option('ct_mode', 'none')
default_clipgrad = self.options['clipgrad'] = self.load_or_def_option('clipgrad', False)
default_pretrain = self.options['pretrain'] = self.load_or_def_option('pretrain', False)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-256", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 16.")
resolution = np.clip ( (resolution // 16) * 16, 64, 256)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['h','mf','f'], help_message="Half / mid face / full face. Half face has better resolution, but covers less area of cheeks. Mid face is 30% wider than half face.").lower()
if (self.is_first_run() or ask_override) and len(device_config.devices) == 1:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
if self.is_first_run():
self.options['archi'] = io.input_str ("AE architecture", default_archi, ['dfhd','liaehd','df','liae'], help_message="'df' keeps faces more natural. 'liae' can fix overly different face shapes. 'hd' is heavyweight version for the best quality.").lower() #-s version is slower, but has decreased change to collapse.
self.options['ae_dims'] = np.clip ( io.input_int("AutoEncoder dimensions", default_ae_dims, add_info="32-1024", help_message="All face information will packed to AE dims. If amount of AE dims are not enough, then for example closed eyes will not be recognized. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 1024 )
e_dims = np.clip ( io.input_int("Encoder dimensions", default_e_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['e_dims'] = e_dims + e_dims % 2
d_dims = np.clip ( io.input_int("Decoder dimensions", default_d_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['d_dims'] = d_dims + d_dims % 2
d_mask_dims = np.clip ( io.input_int("Decoder mask dimensions", default_d_mask_dims, add_info="16-256", help_message="Typical mask dimensions = decoder dimensions / 3. If you manually cut out obstacles from the dst mask, you can increase this parameter to achieve better quality." ), 16, 256 )
self.options['d_mask_dims'] = d_mask_dims + d_mask_dims % 2
if self.is_first_run() or ask_override:
self.options['learn_mask'] = io.input_bool ("Learn mask", default_learn_mask, help_message="Learning mask can help model to recognize face directions. Learn without mask can reduce model size, in this case merger forced to use 'not predicted mask' that is not smooth as predicted.")
self.options['lr_dropout'] = io.input_bool ("Use learning rate dropout", default_lr_dropout, help_message="When the face is trained enough, you can enable this option to get extra sharpness for less amount of iterations.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness for less amount of iterations.")
if 'df' in self.options['archi']:
self.options['true_face_training'] = io.input_bool ("Enable 'true face' training", default_true_face_training, help_message="The result face will be more like src and will get extra sharpness. Enable it for last 10-20k iterations before conversion.")
else:
self.options['true_face_training'] = False
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 to transfer face style details such as light and color conditions. Warning: Enable it only after 10k iters, when predicted face is clear enough to start learn style. Start from 0.1 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 to transfer background around face. This can make face more like dst. Enabling this option increases the chance of model collapse."), 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['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.")
if self.options['pretrain'] and self.get_pretraining_data_path() is None:
raise Exception("pretraining_data_path is not defined")
self.pretrain_just_disabled = (default_pretrain == True and self.options['pretrain'] == False)
if self.pretrain_just_disabled:
self.set_iter(1)
def on_initialize_options(self):
device_config = nn.getCurrentDeviceConfig()
lowest_vram = 2
if len(device_config.devices) != 0:
lowest_vram = device_config.devices.get_worst_device().total_mem_gb
if lowest_vram >= 4:
suggest_batch_size = 8
else:
suggest_batch_size = 4
yn_str = {True:'y',False:'n'}
min_res = 64
max_res = 640
default_resolution = self.options['resolution'] = self.load_or_def_option('resolution', 128)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'f')
default_models_opt_on_gpu = self.options['models_opt_on_gpu'] = self.load_or_def_option('models_opt_on_gpu', True)
archi = self.load_or_def_option('archi', 'df')
archi = {'dfuhd':'df-u','liaeuhd':'liae-u'}.get(archi, archi) #backward comp
default_archi = self.options['archi'] = archi
default_ae_dims = self.options['ae_dims'] = self.load_or_def_option('ae_dims', 256)
default_e_dims = self.options['e_dims'] = self.load_or_def_option('e_dims', 64)
default_d_dims = self.options['d_dims'] = self.options.get('d_dims', None)
default_d_mask_dims = self.options['d_mask_dims'] = self.options.get('d_mask_dims', None)
default_masked_training = self.options['masked_training'] = self.load_or_def_option('masked_training', True)
default_eyes_prio = self.options['eyes_prio'] = self.load_or_def_option('eyes_prio', False)
default_uniform_yaw = self.options['uniform_yaw'] = self.load_or_def_option('uniform_yaw', False)
lr_dropout = self.load_or_def_option('lr_dropout', 'n')
lr_dropout = {True:'y', False:'n'}.get(lr_dropout, lr_dropout) #backward comp
default_lr_dropout = self.options['lr_dropout'] = lr_dropout
default_random_warp = self.options['random_warp'] = self.load_or_def_option('random_warp', True)
default_gan_power = self.options['gan_power'] = self.load_or_def_option('gan_power', 0.0)
default_true_face_power = self.options['true_face_power'] = self.load_or_def_option('true_face_power', 0.0)
default_face_style_power = self.options['face_style_power'] = self.load_or_def_option('face_style_power', 0.0)
default_bg_style_power = self.options['bg_style_power'] = self.load_or_def_option('bg_style_power', 0.0)
default_ct_mode = self.options['ct_mode'] = self.load_or_def_option('ct_mode', 'none')
default_clipgrad = self.options['clipgrad'] = self.load_or_def_option('clipgrad', False)
default_pretrain = self.options['pretrain'] = self.load_or_def_option('pretrain', False)
ask_override = self.ask_override()
if self.is_first_run() or ask_override:
self.ask_autobackup_hour()
self.ask_write_preview_history()
self.ask_target_iter()
self.ask_random_flip()
self.ask_batch_size(suggest_batch_size)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-640", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 16 and 32 for -d archi.")
resolution = np.clip ( (resolution // 16) * 16, min_res, max_res)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['h','mf','f','wf','head'], help_message="Half / mid face / full face / whole face / head. Half face has better resolution, but covers less area of cheeks. Mid face is 30% wider than half face. 'Whole face' covers full area of face include forehead. 'head' covers full head, but requires XSeg for src and dst faceset.").lower()
while True:
archi = io.input_str ("AE architecture", default_archi, help_message=\
"""
'df' keeps more identity-preserved face.
'liae' can fix overly different face shapes.
'-u' increased likeness of the face.
'-d' (experimental) doubling the resolution using the same computation cost.
Examples: df, liae, df-d, df-ud, liae-ud, ...
""").lower()
archi_split = archi.split('-')
if len(archi_split) == 2:
archi_type, archi_opts = archi_split
elif len(archi_split) == 1:
archi_type, archi_opts = archi_split[0], None
else:
continue
if archi_type not in ['df', 'liae']:
continue
if archi_opts is not None:
if len(archi_opts) == 0:
continue
if len([ 1 for opt in archi_opts if opt not in ['u','d'] ]) != 0:
continue
if 'd' in archi_opts:
self.options['resolution'] = np.clip ( (self.options['resolution'] // 32) * 32, min_res, max_res)
break
self.options['archi'] = archi
default_d_dims = self.options['d_dims'] = self.load_or_def_option('d_dims', 64)
default_d_mask_dims = default_d_dims // 3
default_d_mask_dims += default_d_mask_dims % 2
default_d_mask_dims = self.options['d_mask_dims'] = self.load_or_def_option('d_mask_dims', default_d_mask_dims)
if self.is_first_run():
self.options['ae_dims'] = np.clip ( io.input_int("AutoEncoder dimensions", default_ae_dims, add_info="32-1024", help_message="All face information will packed to AE dims. If amount of AE dims are not enough, then for example closed eyes will not be recognized. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 1024 )
e_dims = np.clip ( io.input_int("Encoder dimensions", default_e_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['e_dims'] = e_dims + e_dims % 2
d_dims = np.clip ( io.input_int("Decoder dimensions", default_d_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['d_dims'] = d_dims + d_dims % 2
d_mask_dims = np.clip ( io.input_int("Decoder mask dimensions", default_d_mask_dims, add_info="16-256", help_message="Typical mask dimensions = decoder dimensions / 3. If you manually cut out obstacles from the dst mask, you can increase this parameter to achieve better quality." ), 16, 256 )
self.options['d_mask_dims'] = d_mask_dims + d_mask_dims % 2
if self.is_first_run() or ask_override:
if self.options['face_type'] == 'wf' or self.options['face_type'] == 'head':
self.options['masked_training'] = io.input_bool ("Masked training", default_masked_training, help_message="This option is available only for 'whole_face' or 'head' type. Masked training clips training area to full_face mask or XSeg mask, thus network will train the faces properly.")
self.options['eyes_prio'] = io.input_bool ("Eyes priority", default_eyes_prio, help_message='Helps to fix eye problems during training like "alien eyes" and wrong eyes direction ( especially on HD architectures ) by forcing the neural network to train eyes with higher priority. before/after https://i.imgur.com/YQHOuSR.jpg ')
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.')
if self.is_first_run() or ask_override:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
self.options['lr_dropout'] = io.input_str (f"Use learning rate dropout", default_lr_dropout, ['n','y','cpu'], help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations. Enabled it before `disable random warp` and before GAN. \nn - disabled.\ny - enabled\ncpu - enabled on CPU. This allows not to use extra VRAM, sacrificing 20% time of iteration.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['gan_power'] = np.clip ( io.input_number ("GAN power", default_gan_power, add_info="0.0 .. 10.0", help_message="Train the network in Generative Adversarial manner. Forces the neural network to learn small details of the face. Enable it only when the face is trained enough and don't disable. Typical value is 0.1"), 0.0, 10.0 )
if 'df' in self.options['archi']:
self.options['true_face_power'] = np.clip ( io.input_number ("'True face' power.", default_true_face_power, add_info="0.0000 .. 1.0", help_message="Experimental option. Discriminates result face to be more like src face. Higher value - stronger discrimination. Typical value is 0.01 . Comparison - https://i.imgur.com/czScS9q.png"), 0.0, 1.0 )
else:
self.options['true_face_power'] = 0.0
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['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.")
if self.options['pretrain'] and self.get_pretraining_data_path() is None:
raise Exception("pretraining_data_path is not defined")
self.pretrain_just_disabled = (default_pretrain == True and self.options['pretrain'] == False)
def video_from_sequence_(input_dir,
output_file,
reference_file=None,
ext=None,
fps=None,
bitrate=None,
include_audio=False,
lossless=None):
input_path = Path(input_dir)
output_file_path = Path(output_file)
reference_file_path = Path(
reference_file) if reference_file is not None else None
if not input_path.exists():
io.log_err("input_dir not found.")
return
if not output_file_path.parent.exists():
output_file_path.parent.mkdir(parents=True, exist_ok=True)
return
out_ext = output_file_path.suffix
if ext is None:
ext = io.input_str("Input image format (extension)", "png")
if lossless is None:
lossless = io.input_bool("Use lossless codec", False)
video_id = None
audio_id = None
ref_in_a = None
if reference_file_path is not None:
if reference_file_path.suffix == '.*':
reference_file_path = pathex.get_first_file_by_stem(
reference_file_path.parent, reference_file_path.stem)
else:
if not reference_file_path.exists():
reference_file_path = None
if reference_file_path is None:
io.log_err("reference_file not found.")
return
#probing reference file
probe = ffmpeg.probe(str(reference_file_path))
#getting first video and audio streams id with fps
for stream in probe['streams']:
if video_id is None and stream['codec_type'] == 'video':
video_id = stream['index']
fps = stream['r_frame_rate']
if audio_id is None and stream['codec_type'] == 'audio':
audio_id = stream['index']
if audio_id is not None:
#has audio track
ref_in_a = ffmpeg.input(str(reference_file_path))[str(audio_id)]
if fps is None:
#if fps not specified and not overwritten by reference-file
fps = max(1, io.input_int("Enter FPS", 25))
if not lossless and bitrate is None:
bitrate = 1 #max (1, settings.bitrate)#io.input_int ("Bitrate of output file in MB/s", 16)
input_image_paths = pathex.get_image_paths(input_path)
i_in = ffmpeg.input('pipe:', format='image2pipe', r=fps)
output_args = [i_in]
if include_audio and ref_in_a is not None:
output_args += [ref_in_a]
output_args += [str(output_file_path)]
output_kwargs = {}
if lossless:
output_kwargs.update({
"c:v": "libx264",
"crf": "0",
"pix_fmt": "yuv420p",
})
else:
output_kwargs.update({
"c:v": "libx264",
"b:v": "%dM" % (1),
"pix_fmt": "yuv420p",
})
if include_audio and ref_in_a is not None:
output_kwargs.update({
"c:a": "aac",
"b:a": "192k",
"ar": "48000",
"strict": "experimental"
})
job = (ffmpeg.output(*output_args, **output_kwargs).overwrite_output())
try:
job_run = job.run_async(pipe_stdin=True)
for image_path in input_image_paths:
with open(image_path, "rb") as f:
image_bytes = f.read()
job_run.stdin.write(image_bytes)
job_run.stdin.close()
job_run.wait()
except:
io.log_err("ffmpeg fail, job commandline:" + str(job.compile()))
def on_initialize_options(self):
device_config = nn.getCurrentDeviceConfig()
lowest_vram = 2
if len(device_config.devices) != 0:
lowest_vram = device_config.devices.get_worst_device().total_mem_gb
if lowest_vram >= 4:
suggest_batch_size = 8
else:
suggest_batch_size = 4
yn_str = {True:'y',False:'n'}
default_resolution = self.options['resolution'] = self.load_or_def_option('resolution', 128)
default_face_type = self.options['face_type'] = self.load_or_def_option('face_type', 'f')
default_models_opt_on_gpu = self.options['models_opt_on_gpu'] = self.load_or_def_option('models_opt_on_gpu', True)
default_archi = self.options['archi'] = self.load_or_def_option('archi', 'df')
default_ae_dims = self.options['ae_dims'] = self.load_or_def_option('ae_dims', 256)
default_e_dims = self.options['e_dims'] = self.load_or_def_option('e_dims', 64)
default_d_dims = self.options['d_dims'] = self.options.get('d_dims', None)
default_d_mask_dims = self.options['d_mask_dims'] = self.options.get('d_mask_dims', None)
default_masked_training = self.options['masked_training'] = self.load_or_def_option('masked_training', True)
default_eyes_prio = self.options['eyes_prio'] = self.load_or_def_option('eyes_prio', False)
default_lr_dropout = self.options['lr_dropout'] = self.load_or_def_option('lr_dropout', False)
default_random_warp = self.options['random_warp'] = self.load_or_def_option('random_warp', True)
default_gan_power = self.options['gan_power'] = self.load_or_def_option('gan_power', 0.0)
default_true_face_power = self.options['true_face_power'] = self.load_or_def_option('true_face_power', 0.0)
default_face_style_power = self.options['face_style_power'] = self.load_or_def_option('face_style_power', 0.0)
default_bg_style_power = self.options['bg_style_power'] = self.load_or_def_option('bg_style_power', 0.0)
default_ct_mode = self.options['ct_mode'] = self.load_or_def_option('ct_mode', 'none')
default_clipgrad = self.options['clipgrad'] = self.load_or_def_option('clipgrad', False)
default_pretrain = self.options['pretrain'] = self.load_or_def_option('pretrain', False)
ask_override = self.ask_override()
if self.is_first_run() or ask_override:
self.ask_autobackup_hour()
self.ask_write_preview_history()
self.ask_target_iter()
self.ask_random_flip()
self.ask_batch_size(suggest_batch_size)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-512", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 16.")
resolution = np.clip ( (resolution // 16) * 16, 64, 512)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['h','mf','f','wf'], help_message="Half / mid face / full face / whole face. Half face has better resolution, but covers less area of cheeks. Mid face is 30% wider than half face. 'Whole face' covers full area of face include forehead, but requires manual merge in Adobe After Effects.").lower()
self.options['archi'] = io.input_str ("AE architecture", default_archi, ['df','liae','dfhd','liaehd'], help_message="'df' keeps faces more natural.\n'liae' can fix overly different face shapes.\n'hd' are experimental versions.").lower()
default_d_dims = 48 if self.options['archi'] == 'dfhd' else 64
default_d_dims = self.options['d_dims'] = self.load_or_def_option('d_dims', default_d_dims)
default_d_mask_dims = default_d_dims // 3
default_d_mask_dims += default_d_mask_dims % 2
default_d_mask_dims = self.options['d_mask_dims'] = self.load_or_def_option('d_mask_dims', default_d_mask_dims)
if self.is_first_run():
self.options['ae_dims'] = np.clip ( io.input_int("AutoEncoder dimensions", default_ae_dims, add_info="32-1024", help_message="All face information will packed to AE dims. If amount of AE dims are not enough, then for example closed eyes will not be recognized. More dims are better, but require more VRAM. You can fine-tune model size to fit your GPU." ), 32, 1024 )
e_dims = np.clip ( io.input_int("Encoder dimensions", default_e_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['e_dims'] = e_dims + e_dims % 2
d_dims = np.clip ( io.input_int("Decoder dimensions", default_d_dims, add_info="16-256", help_message="More dims help to recognize more facial features and achieve sharper result, but require more VRAM. You can fine-tune model size to fit your GPU." ), 16, 256 )
self.options['d_dims'] = d_dims + d_dims % 2
d_mask_dims = np.clip ( io.input_int("Decoder mask dimensions", default_d_mask_dims, add_info="16-256", help_message="Typical mask dimensions = decoder dimensions / 3. If you manually cut out obstacles from the dst mask, you can increase this parameter to achieve better quality." ), 16, 256 )
self.options['d_mask_dims'] = d_mask_dims + d_mask_dims % 2
if self.is_first_run() or ask_override:
if self.options['face_type'] == 'wf':
self.options['masked_training'] = io.input_bool ("Masked training", default_masked_training, help_message="This option is available only for 'whole_face' type. Masked training clips training area to full_face mask, thus network will train the faces properly. When the face is trained enough, disable this option to train all area of the frame. Merge with 'raw-rgb' mode, then use Adobe After Effects to manually mask and compose whole face include forehead.")
self.options['eyes_prio'] = io.input_bool ("Eyes priority", default_eyes_prio, help_message='Helps to fix eye problems during training like "alien eyes" and wrong eyes direction ( especially on HD architectures ) by forcing the neural network to train eyes with higher priority. before/after https://i.imgur.com/YQHOuSR.jpg ')
if self.is_first_run() or ask_override:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
self.options['lr_dropout'] = io.input_bool ("Use learning rate dropout", default_lr_dropout, help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['gan_power'] = np.clip ( io.input_number ("GAN power", default_gan_power, add_info="0.0 .. 10.0", help_message="Train the network in Generative Adversarial manner. Accelerates the speed of training. Forces the neural network to learn small details of the face. You can enable/disable this option at any time. Typical value is 1.0"), 0.0, 10.0 )
if 'df' in self.options['archi']:
self.options['true_face_power'] = np.clip ( io.input_number ("'True face' power.", default_true_face_power, add_info="0.0000 .. 1.0", help_message="Experimental option. Discriminates result face to be more like src face. Higher value - stronger discrimination. Typical value is 0.01 . Comparison - https://i.imgur.com/czScS9q.png"), 0.0, 1.0 )
else:
self.options['true_face_power'] = 0.0
if self.options['face_type'] != 'wf':
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 to transfer face style details such as light and color conditions. 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 to transfer background around face. 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['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.")
if self.options['pretrain'] and self.get_pretraining_data_path() is None:
raise Exception("pretraining_data_path is not defined")
self.pretrain_just_disabled = (default_pretrain == True and self.options['pretrain'] == False)
