Upload 4 files

processors/modules/expression_restorer/choices.py

@@ -0,0 +1,10 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_int_range
+from facefusion.processors.modules.expression_restorer.types import ExpressionRestorerArea, ExpressionRestorerModel
+
+expression_restorer_models : List[ExpressionRestorerModel] = list(get_args(ExpressionRestorerModel))
+
+expression_restorer_areas : List[ExpressionRestorerArea] = list(get_args(ExpressionRestorerArea))
+
+expression_restorer_factor_range : Sequence[int] = create_int_range(0, 100, 1)

processors/modules/expression_restorer/core.py

@@ -0,0 +1,280 @@
+from argparse import ArgumentParser
+from functools import lru_cache
+from types import ModuleType
+from typing import List, Tuple
+
+import cv2
+import numpy
+
+import facefusion.jobs.job_manager
+import facefusion.jobs.job_store
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_int_metavar, get_middle
+from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
+from facefusion.face_creator import scale_face
+from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
+from facefusion.face_masker import create_box_mask, create_occlusion_mask
+from facefusion.face_selector import select_faces
+from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
+from facefusion.processors.live_portrait import create_rotation, limit_expression
+from facefusion.processors.modules.expression_restorer import choices as expression_restorer_choices
+from facefusion.processors.modules.expression_restorer.types import ExpressionRestorerInputs
+from facefusion.processors.types import LivePortraitExpression, LivePortraitFeatureVolume, LivePortraitMotionPoints, LivePortraitPitch, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitYaw, ProcessorOutputs
+from facefusion.program_helper import find_argument_group
+from facefusion.thread_helper import conditional_thread_semaphore, thread_semaphore
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_video_frame
+
+
+@lru_cache()
+def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
+	return\
+	{
+		'live_portrait':
+		{
+			'__metadata__':
+			{
+				'vendor': 'KwaiVGI',
+				'license': 'MIT',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'feature_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.hash')
+				},
+				'motion_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.hash')
+				},
+				'generator':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_generator.hash')
+				}
+			},
+			'sources':
+			{
+				'feature_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.onnx')
+				},
+				'motion_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.onnx')
+				},
+				'generator':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_generator.onnx')
+				}
+			},
+			'template': 'arcface_128',
+			'size': (512, 512)
+		}
+	}
+
+
+def get_inference_pool() -> InferencePool:
+	model_names = [ state_manager.get_item('expression_restorer_model') ]
+	model_source_set = get_model_options().get('sources')
+
+	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
+
+
+def clear_inference_pool() -> None:
+	model_names = [ state_manager.get_item('expression_restorer_model') ]
+	inference_manager.clear_inference_pool(__name__, model_names)
+
+
+def get_model_options() -> ModelOptions:
+	model_name = state_manager.get_item('expression_restorer_model')
+	return create_static_model_set('full').get(model_name)
+
+
+def register_args(program : ArgumentParser) -> None:
+	group_processors = find_argument_group(program, 'processors')
+	if group_processors:
+		group_processors.add_argument('--expression-restorer-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'expression_restorer_model', 'live_portrait'), choices = expression_restorer_choices.expression_restorer_models)
+		group_processors.add_argument('--expression-restorer-factor', help = translator.get('help.factor', __package__), type = int, default = config.get_int_value('processors', 'expression_restorer_factor', '80'), choices = expression_restorer_choices.expression_restorer_factor_range, metavar = create_int_metavar(expression_restorer_choices.expression_restorer_factor_range))
+		group_processors.add_argument('--expression-restorer-areas', help = translator.get('help.areas', __package__).format(choices = ', '.join(expression_restorer_choices.expression_restorer_areas)), default = config.get_str_list('processors', 'expression_restorer_areas', ' '.join(expression_restorer_choices.expression_restorer_areas)), choices = expression_restorer_choices.expression_restorer_areas, nargs = '+', metavar = 'EXPRESSION_RESTORER_AREAS')
+		facefusion.jobs.job_store.register_step_keys([ 'expression_restorer_model', 'expression_restorer_factor', 'expression_restorer_areas' ])
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('expression_restorer_model', args.get('expression_restorer_model'))
+	apply_state_item('expression_restorer_factor', args.get('expression_restorer_factor'))
+	apply_state_item('expression_restorer_areas', args.get('expression_restorer_areas'))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
+def pre_check() -> bool:
+	model_hash_set = get_model_options().get('hashes')
+	model_source_set = get_model_options().get('sources')
+
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
+	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
+
+
+def pre_process(mode : ProcessMode) -> bool:
+	if mode == 'stream':
+		logger.error(translator.get('stream_not_supported') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
+		return False
+	return True
+
+
+def post_process() -> None:
+	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
+	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
+		clear_inference_pool()
+
+	if state_manager.get_item('video_memory_strategy') == 'strict':
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
+
+
+def restore_expression(target_face : Face, target_vision_frame : VisionFrame, temp_vision_frame : VisionFrame) -> VisionFrame:
+	model_template = get_model_options().get('template')
+	model_size = get_model_options().get('size')
+	expression_restorer_factor = float(numpy.interp(float(state_manager.get_item('expression_restorer_factor')), [ 0, 100 ], [ 0, 1.2 ]))
+	target_crop_vision_frame, _ = warp_face_by_face_landmark_5(target_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
+	temp_crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
+	box_mask = create_box_mask(temp_crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+	crop_masks =\
+	[
+		box_mask
+	]
+
+	if 'occlusion' in state_manager.get_item('face_mask_types'):
+		occlusion_mask = create_occlusion_mask(temp_crop_vision_frame)
+		crop_masks.append(occlusion_mask)
+
+	target_crop_vision_frame = prepare_crop_frame(target_crop_vision_frame)
+	temp_crop_vision_frame = prepare_crop_frame(temp_crop_vision_frame)
+	temp_crop_vision_frame = apply_restore(target_crop_vision_frame, temp_crop_vision_frame, expression_restorer_factor)
+	temp_crop_vision_frame = normalize_crop_frame(temp_crop_vision_frame)
+	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+	paste_vision_frame = paste_back(temp_vision_frame, temp_crop_vision_frame, crop_mask, affine_matrix)
+	return paste_vision_frame
+
+
+def apply_restore(target_crop_vision_frame : VisionFrame, temp_crop_vision_frame : VisionFrame, expression_restorer_factor : float) -> VisionFrame:
+	feature_volume = forward_extract_feature(temp_crop_vision_frame)
+	target_expression = forward_extract_motion(target_crop_vision_frame)[5]
+	pitch, yaw, roll, scale, translation, temp_expression, motion_points = forward_extract_motion(temp_crop_vision_frame)
+	rotation = create_rotation(pitch, yaw, roll)
+	target_expression = restrict_expression_areas(temp_expression, target_expression)
+	target_expression = target_expression * expression_restorer_factor + temp_expression * (1 - expression_restorer_factor)
+	target_expression = limit_expression(target_expression)
+	target_motion_points = scale * (motion_points @ rotation.T + target_expression) + translation
+	temp_motion_points = scale * (motion_points @ rotation.T + temp_expression) + translation
+	crop_vision_frame = forward_generate_frame(feature_volume, target_motion_points, temp_motion_points)
+	return crop_vision_frame
+
+
+def restrict_expression_areas(temp_expression : LivePortraitExpression, target_expression : LivePortraitExpression) -> LivePortraitExpression:
+	expression_restorer_areas = state_manager.get_item('expression_restorer_areas')
+
+	if 'upper-face' not in expression_restorer_areas:
+		target_expression[:, [ 1, 2, 6, 10, 11, 12, 13, 15, 16 ]] = temp_expression[:, [ 1, 2, 6, 10, 11, 12, 13, 15, 16 ]]
+
+	if 'lower-face' not in expression_restorer_areas:
+		target_expression[:, [ 3, 7, 14, 17, 18, 19, 20 ]] = temp_expression[:, [ 3, 7, 14, 17, 18, 19, 20 ]]
+
+	target_expression[:, [ 0, 4, 5, 8, 9 ]] = temp_expression[:, [ 0, 4, 5, 8, 9 ]]
+	return target_expression
+
+
+def forward_extract_feature(crop_vision_frame : VisionFrame) -> LivePortraitFeatureVolume:
+	feature_extractor = get_inference_pool().get('feature_extractor')
+
+	with conditional_thread_semaphore():
+		feature_volume = feature_extractor.run(None,
+		{
+			'input': crop_vision_frame
+		})[0]
+
+	return feature_volume
+
+
+def forward_extract_motion(crop_vision_frame : VisionFrame) -> Tuple[LivePortraitPitch, LivePortraitYaw, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitExpression, LivePortraitMotionPoints]:
+	motion_extractor = get_inference_pool().get('motion_extractor')
+
+	with conditional_thread_semaphore():
+		pitch, yaw, roll, scale, translation, expression, motion_points = motion_extractor.run(None,
+		{
+			'input': crop_vision_frame
+		})
+
+	return pitch, yaw, roll, scale, translation, expression, motion_points
+
+
+def forward_generate_frame(feature_volume : LivePortraitFeatureVolume, target_motion_points : LivePortraitMotionPoints, temp_motion_points : LivePortraitMotionPoints) -> VisionFrame:
+	generator = get_inference_pool().get('generator')
+
+	with thread_semaphore():
+		crop_vision_frame = generator.run(None,
+		{
+			'feature_volume': feature_volume,
+			'source': target_motion_points,
+			'target': temp_motion_points
+		})[0][0]
+
+	return crop_vision_frame
+
+
+def prepare_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
+	model_size = get_model_options().get('size')
+	prepare_size = (model_size[0] // 2, model_size[1] // 2)
+	crop_vision_frame = cv2.resize(crop_vision_frame, prepare_size, interpolation = cv2.INTER_AREA)
+	crop_vision_frame = crop_vision_frame[:, :, ::-1] / 255.0
+	crop_vision_frame = numpy.expand_dims(crop_vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
+	return crop_vision_frame
+
+
+def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
+	crop_vision_frame = crop_vision_frame.transpose(1, 2, 0).clip(0, 1)
+	crop_vision_frame = crop_vision_frame * 255.0
+	crop_vision_frame = crop_vision_frame.astype(numpy.uint8)[:, :, ::-1]
+	return crop_vision_frame
+
+
+def process_frame(inputs : ExpressionRestorerInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
+
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
+
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = restore_expression(target_face, target_vision_frame, temp_vision_frame)
+
+	return temp_vision_frame, temp_vision_mask

processors/modules/expression_restorer/locales.py

@@ -0,0 +1,20 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for restoring the expression',
+			'factor': 'restore factor of expression from the target face',
+			'areas': 'choose the items used for the expression areas (choices: {choices})'
+		},
+		'uis':
+		{
+			'model_dropdown': 'EXPRESSION RESTORER MODEL',
+			'factor_slider': 'EXPRESSION RESTORER FACTOR',
+			'areas_checkbox_group': 'EXPRESSION RESTORER AREAS'
+		}
+	}
+}

processors/modules/expression_restorer/types.py

@@ -0,0 +1,16 @@
+from typing import List, Literal, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+ExpressionRestorerInputs = TypedDict('ExpressionRestorerInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+ExpressionRestorerModel = Literal['live_portrait']
+
+ExpressionRestorerArea = Literal['upper-face', 'lower-face']