merge stuff

VIOLATION.md

@@ -1,60 +0,0 @@
-# Naming Convention Violations
-
-## Variables using `xxx_frame` instead of `xxx_vision_frame`
-
-- `source_frame_resize` → `source_vision_frame_resize` in `facefusion/vision.py:290`
-- `target_frame_resize` → `target_vision_frame_resize` in `facefusion/vision.py:291`
-
-## Variables using `xxx_mask` instead of `xxx_vision_mask`
-
-### box_mask → box_vision_mask
-
-- `facefusion/face_masker.py:192`
-- `facefusion/processors/modules/face_debugger/core.py:131`
-- `facefusion/processors/modules/expression_restorer/core.py:158`
-- `facefusion/processors/modules/age_modifier/core.py:179,205`
-- `facefusion/processors/modules/face_enhancer/core.py:346`
-- `facefusion/processors/modules/face_editor/core.py:208`
-- `facefusion/processors/modules/deep_swapper/core.py:331`
-- `facefusion/processors/modules/face_swapper/core.py:591`
-- `facefusion/processors/modules/lip_syncer/core.py:189`
-
-### occlusion_mask → occlusion_vision_mask
-
-- `facefusion/face_masker.py:221,222`
-- `facefusion/processors/modules/face_debugger/core.py:135`
-- `facefusion/processors/modules/expression_restorer/core.py:165`
-- `facefusion/processors/modules/age_modifier/core.py:186,212,214`
-- `facefusion/processors/modules/face_enhancer/core.py:353`
-- `facefusion/processors/modules/deep_swapper/core.py:338`
-- `facefusion/processors/modules/face_swapper/core.py:595`
-- `facefusion/processors/modules/lip_syncer/core.py:184`
-
-### area_mask → area_vision_mask
-
-- `facefusion/face_masker.py:235,237`
-- `facefusion/processors/modules/face_debugger/core.py:140`
-- `facefusion/processors/modules/deep_swapper/core.py:350`
-- `facefusion/processors/modules/face_swapper/core.py:608`
-- `facefusion/processors/modules/lip_syncer/core.py:197`
-
-### region_mask → region_vision_mask
-
-- `facefusion/face_masker.py:250,252,253`
-- `facefusion/processors/modules/face_debugger/core.py:144`
-- `facefusion/processors/modules/deep_swapper/core.py:354`
-- `facefusion/processors/modules/face_swapper/core.py:612`
-
-### crop_mask → crop_vision_mask
-
-- `facefusion/processors/modules/face_debugger/core.py:147,148,149`
-- `facefusion/processors/modules/expression_restorer/core.py:172,173`
-- `facefusion/processors/modules/age_modifier/core.py:197,224,225`
-- `facefusion/processors/modules/face_enhancer/core.py:360,361`
-- `facefusion/processors/modules/deep_swapper/core.py:357,405,406,407,408`
-- `facefusion/processors/modules/face_swapper/core.py:615,616`
-- `facefusion/processors/modules/lip_syncer/core.py:206,207`
-
-### temp_mask → temp_vision_mask
-
-- `facefusion/face_masker.py:216,217,218`

facefusion/processors/modules/age_modifier/core.py

@@ -29,6 +29,41 @@ from facefusion.vision import match_frame_color, read_static_image, read_static_
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
+		'fran':
+		{
+			'__metadata__':
+			{
+				'vendor': 'ry-lu',
+				'license': 'mit',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.6.0', 'fran.hash'),
+					'path': resolve_relative_path('../.assets/models/fran.hash')
+				}
+			},
+			'sources':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.6.0', 'fran.onnx'),
+					'path': resolve_relative_path('../.assets/models/fran.onnx')
+				}
+			},
+			'templates':
+			{
+				'target': 'ffhq_512',
+			},
+			'sizes':
+			{
+				'target': (1024, 1024),
+			},
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
 		'styleganex_age':
 		{
 			'__metadata__':
@@ -62,7 +97,9 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			{
 				'target': (256, 256),
 				'target_with_background': (384, 384)
-			}
+			},
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
 		}
 	}
 
@@ -92,7 +129,7 @@ def register_args(program : ArgumentParser) -> None:
 				group_processors.add_argument(
 					'--age-modifier-model',
 					help = translator.get('help.model', __package__),
-					default = config.get_str_value('processors', 'age_modifier_model', 'styleganex_age'),
+					default = config.get_str_value('processors', 'age_modifier_model', 'fran'),
 					choices = age_modifier_choices.age_modifier_models
 				),
 				group_processors.add_argument(
@@ -151,32 +188,57 @@ def modify_age(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFra
 	model_sizes = get_model_options().get('sizes')
 	face_landmark_5 = target_face.landmark_set.get('5/68').copy()
 	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5, model_templates.get('target'), model_sizes.get('target'))
-	extend_face_landmark_5 = scale_face_landmark_5(face_landmark_5, 0.875)
-	extend_vision_frame, extend_affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, extend_face_landmark_5, model_templates.get('target_with_background'), model_sizes.get('target_with_background'))
-	extend_vision_frame_raw = extend_vision_frame.copy()
-	box_mask = create_box_mask(extend_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(crop_vision_frame)
-		temp_matrix = merge_matrix([ extend_affine_matrix, cv2.invertAffineTransform(affine_matrix) ])
-		occlusion_mask = cv2.warpAffine(occlusion_mask, temp_matrix, model_sizes.get('target_with_background'))
-		crop_masks.append(occlusion_mask)
+	if state_manager.get_item('age_modifier_model') == 'fran':
+		box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+		crop_masks =\
+		[
+			box_mask
+		]
 
-	crop_vision_frame = prepare_vision_frame(crop_vision_frame)
-	extend_vision_frame = prepare_vision_frame(extend_vision_frame)
-	age_modifier_direction = numpy.array(numpy.interp(state_manager.get_item('age_modifier_direction'), [ -100, 100 ], [ 2.5, -2.5 ])).astype(numpy.float32)
-	extend_vision_frame = forward(crop_vision_frame, extend_vision_frame, age_modifier_direction)
-	extend_vision_frame = normalize_extend_frame(extend_vision_frame)
-	extend_vision_frame = match_frame_color(extend_vision_frame_raw, extend_vision_frame)
-	extend_affine_matrix *= (model_sizes.get('target')[0] * 4) / model_sizes.get('target_with_background')[0]
-	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
-	crop_mask = cv2.resize(crop_mask, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4))
-	paste_vision_frame = paste_back(temp_vision_frame, extend_vision_frame, crop_mask, extend_affine_matrix)
-	return paste_vision_frame
+		if 'occlusion' in state_manager.get_item('face_mask_types'):
+			occlusion_mask = create_occlusion_mask(crop_vision_frame)
+			crop_masks.append(occlusion_mask)
+
+		crop_vision_frame = prepare_vision_frame(crop_vision_frame)
+		target_age = numpy.mean(target_face.age)
+		age_modifier_direction = numpy.array([ target_age, target_age + state_manager.get_item('age_modifier_direction') ], dtype = numpy.float32) / 100
+		age_modifier_direction = age_modifier_direction.clip(0, 1)
+		crop_vision_frame = forward(crop_vision_frame, crop_vision_frame, age_modifier_direction)
+		crop_vision_frame = normalize_vision_frame(crop_vision_frame)
+		crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+		paste_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, crop_mask, affine_matrix)
+		return paste_vision_frame
+
+	if state_manager.get_item('age_modifier_model') == 'styleganex_age':
+		extend_face_landmark_5 = scale_face_landmark_5(face_landmark_5, 0.875)
+		extend_vision_frame, extend_affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, extend_face_landmark_5, model_templates.get('target_with_background'), model_sizes.get('target_with_background'))
+		extend_vision_frame_raw = extend_vision_frame.copy()
+		box_mask = create_box_mask(extend_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(crop_vision_frame)
+			temp_matrix = merge_matrix([ extend_affine_matrix, cv2.invertAffineTransform(affine_matrix) ])
+			occlusion_mask = cv2.warpAffine(occlusion_mask, temp_matrix, model_sizes.get('target_with_background'))
+			crop_masks.append(occlusion_mask)
+
+		crop_vision_frame = prepare_vision_frame(crop_vision_frame)
+		extend_vision_frame = prepare_vision_frame(extend_vision_frame)
+		age_modifier_direction = numpy.array(numpy.interp(state_manager.get_item('age_modifier_direction'), [ -100, 100 ], [ 2.5, -2.5 ])).astype(numpy.float32)
+		extend_vision_frame = forward(crop_vision_frame, extend_vision_frame, age_modifier_direction)
+		extend_vision_frame = normalize_extend_frame(extend_vision_frame)
+		extend_vision_frame = match_frame_color(extend_vision_frame_raw, extend_vision_frame)
+		extend_affine_matrix *= (model_sizes.get('target')[0] * 4) / model_sizes.get('target_with_background')[0]
+		crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+		crop_mask = cv2.resize(crop_mask, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4))
+		paste_vision_frame = paste_back(temp_vision_frame, extend_vision_frame, crop_mask, extend_affine_matrix)
+		return paste_vision_frame
+
+	return temp_vision_frame
 
 
 def forward(crop_vision_frame : VisionFrame, extend_vision_frame : VisionFrame, age_modifier_direction : AgeModifierDirection) -> VisionFrame:
@@ -201,12 +263,24 @@ def forward(crop_vision_frame : VisionFrame, extend_vision_frame : VisionFrame,
 
 
 def prepare_vision_frame(vision_frame : VisionFrame) -> VisionFrame:
+	model_mean = get_model_options().get('mean')
+	model_standard_deviation = get_model_options().get('standard_deviation')
 	vision_frame = vision_frame[:, :, ::-1] / 255.0
-	vision_frame = (vision_frame - 0.5) / 0.5
+	vision_frame = (vision_frame - model_mean) / model_standard_deviation
 	vision_frame = numpy.expand_dims(vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
 	return vision_frame
 
 
+def normalize_vision_frame(vision_frame : VisionFrame) -> VisionFrame:
+	model_mean = get_model_options().get('mean')
+	model_standard_deviation = get_model_options().get('standard_deviation')
+	vision_frame = vision_frame.transpose(1, 2, 0)
+	vision_frame = vision_frame * model_standard_deviation + model_mean
+	vision_frame = vision_frame.clip(0, 1)
+	vision_frame = vision_frame[:, :, ::-1] * 255
+	return vision_frame
+
+
 def normalize_extend_frame(extend_vision_frame : VisionFrame) -> VisionFrame:
 	model_sizes = get_model_options().get('sizes')
 	extend_vision_frame = numpy.clip(extend_vision_frame, -1, 1)

facefusion/processors/modules/background_remover/core.py

@@ -8,7 +8,7 @@ import numpy
 import facefusion.capability_store
 import facefusion.jobs.job_manager
 from facefusion import config, content_analyser, inference_manager, logger, state_manager, translator, video_manager
-from facefusion.common_helper import is_macos
+from facefusion.common_helper import is_macos, is_windows
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
 from facefusion.execution import has_execution_provider
 from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path
@@ -51,6 +51,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/ben_2.onnx')
 				}
 			},
+			'type': 'ben',
 			'size': (1024, 1024),
 			'mean': [ 0.0, 0.0, 0.0 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
@@ -79,6 +80,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/birefnet_general.onnx')
 				}
 			},
+			'type': 'birefnet',
 			'size': (1024, 1024),
 			'mean': [ 0.0, 0.0, 0.0 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
@@ -107,10 +109,69 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/birefnet_portrait.onnx')
 				}
 			},
+			'type': 'birefnet',
 			'size': (1024, 1024),
 			'mean': [ 0.0, 0.0, 0.0 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
 		},
+		'corridor_key_1024':
+		{
+			'__metadata__':
+			{
+				'vendor': 'nikopueringer',
+				'license': 'Non-Commercial',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_1024.hash'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_1024.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_1024.onnx'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_1024.onnx')
+				}
+			},
+			'type': 'corridor_key',
+			'size': (1024, 1024),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'corridor_key_2048':
+		{
+			'__metadata__':
+			{
+				'vendor': 'nikopueringer',
+				'license': 'Non-Commercial',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_2048.hash'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_2048.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_2048.onnx'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_2048.onnx')
+				}
+			},
+			'type': 'corridor_key',
+			'size': (2048, 2048),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
 		'isnet_general':
 		{
 			'__metadata__':
@@ -135,6 +196,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/isnet_general.onnx')
 				}
 			},
+			'type': 'isnet',
 			'size': (1024, 1024),
 			'mean': [ 0.5, 0.5, 0.5 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
@@ -163,6 +225,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/modnet.onnx')
 				}
 			},
+			'type': 'modnet',
 			'size': (512, 512),
 			'mean': [ 0.5, 0.5, 0.5 ],
 			'standard_deviation': [ 0.5, 0.5, 0.5 ]
@@ -191,6 +254,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/ormbg.onnx')
 				}
 			},
+			'type': 'ormbg',
 			'size': (1024, 1024),
 			'mean': [ 0.0, 0.0, 0.0 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
@@ -219,6 +283,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/rmbg_1.4.onnx')
 				}
 			},
+			'type': 'rmbg',
 			'size': (1024, 1024),
 			'mean': [ 0.5, 0.5, 0.5 ],
 			'standard_deviation': [ 1.0, 1.0, 1.0 ]
@@ -247,6 +312,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/rmbg_2.0.onnx')
 				}
 			},
+			'type': 'rmbg',
 			'size': (1024, 1024),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -275,6 +341,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/silueta.onnx')
 				}
 			},
+			'type': 'silueta',
 			'size': (320, 320),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -303,6 +370,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/u2net_cloth.onnx')
 				}
 			},
+			'type': 'u2net_cloth',
 			'size': (768, 768),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -331,6 +399,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/u2net_general.onnx')
 				}
 			},
+			'type': 'u2net',
 			'size': (320, 320),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -359,6 +428,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/u2net_human.onnx')
 				}
 			},
+			'type': 'u2net',
 			'size': (320, 320),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -387,6 +457,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/u2netp.onnx')
 				}
 			},
+			'type': 'u2netp',
 			'size': (320, 320),
 			'mean': [ 0.485, 0.456, 0.406 ],
 			'standard_deviation': [ 0.229, 0.224, 0.225 ]
@@ -407,7 +478,9 @@ def clear_inference_pool() -> None:
 
 
 def resolve_execution_providers() -> List[ExecutionProvider]:
-	if is_macos() and has_execution_provider('coreml'):
+	model_type = get_model_options().get('type')
+
+	if is_macos() and has_execution_provider('coreml') or is_windows() and has_execution_provider('directml') and model_type == 'corridor_key':
 		return [ 'cpu' ]
 	return state_manager.get_item('execution_providers')
 
@@ -425,14 +498,21 @@ def register_args(program : ArgumentParser) -> None:
 				group_processors.add_argument(
 					'--background-remover-model',
 					help = translator.get('help.model', __package__),
-					default = config.get_str_value('processors', 'background_remover_model', 'rmbg_2.0'),
+					default = config.get_str_value('processors', 'background_remover_model', 'modnet'),
 					choices = background_remover_choices.background_remover_models
 				),
 				group_processors.add_argument(
-					'--background-remover-color',
-					help = translator.get('help.color', __package__),
+					'--background-remover-fill-color',
+					help = translator.get('help.fill_color', __package__),
 					type = partial(sanitize_int_range, int_range = background_remover_choices.background_remover_color_range),
-					default = config.get_int_list('processors', 'background_remover_color', '0 0 0 0'),
+					default = config.get_int_list('processors', 'background_remover_fill_color', '0 0 0 0'),
+					nargs = '+'
+				),
+				group_processors.add_argument(
+					'--background-remover-despill-color',
+					help = translator.get('help.despill_color', __package__),
+					type = partial(sanitize_int_range, int_range = background_remover_choices.background_remover_color_range),
+					default = config.get_int_list('processors', 'background_remover_despill_color', '0 0 0 0'),
 					nargs = '+'
 				)
 			],
@@ -442,7 +522,8 @@ def register_args(program : ArgumentParser) -> None:
 
 def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
 	apply_state_item('background_remover_model', args.get('background_remover_model'))
-	apply_state_item('background_remover_color', normalize_color(args.get('background_remover_color')))
+	apply_state_item('background_remover_fill_color', normalize_color(args.get('background_remover_fill_color')))
+	apply_state_item('background_remover_despill_color', normalize_color(args.get('background_remover_despill_color')))
 
 
 def pre_check() -> bool:
@@ -474,16 +555,26 @@ def post_process() -> None:
 
 
 def remove_background(temp_vision_frame : VisionFrame) -> Tuple[VisionFrame, Mask]:
-	temp_vision_mask = forward(prepare_temp_frame(temp_vision_frame))
-	temp_vision_mask = normalize_vision_mask(temp_vision_mask)
-	temp_vision_mask = cv2.resize(temp_vision_mask, temp_vision_frame.shape[:2][::-1])
-	temp_vision_frame = apply_background_color(temp_vision_frame, temp_vision_mask)
-	return temp_vision_frame, temp_vision_mask
+	model_type = get_model_options().get('type')
+
+	if model_type == 'corridor_key':
+		remove_vision_mask, remove_vision_frame = forward_corridor_key(prepare_temp_frame(temp_vision_frame))
+		remove_vision_frame = numpy.squeeze(remove_vision_frame).transpose(1, 2, 0)
+		remove_vision_frame = numpy.clip(remove_vision_frame * 255, 0, 255).astype(numpy.uint8)
+		temp_vision_frame = cv2.resize(remove_vision_frame[:, :, ::-1], temp_vision_frame.shape[:2][::-1])
+	else:
+		remove_vision_mask = forward(prepare_temp_frame(temp_vision_frame))
+
+	remove_vision_mask = normalize_vision_mask(remove_vision_mask)
+	remove_vision_mask = cv2.resize(remove_vision_mask, temp_vision_frame.shape[:2][::-1])
+	temp_vision_frame = apply_despill_color(temp_vision_frame)
+	temp_vision_frame = apply_fill_color(temp_vision_frame, remove_vision_mask)
+	return temp_vision_frame, remove_vision_mask
 
 
 def forward(temp_vision_frame : VisionFrame) -> VisionFrame:
 	background_remover = get_inference_pool().get('background_remover')
-	model_name = state_manager.get_item('background_remover_model')
+	model_type = get_model_options().get('type')
 
 	with thread_semaphore():
 		remove_vision_frame = background_remover.run(None,
@@ -491,20 +582,42 @@ def forward(temp_vision_frame : VisionFrame) -> VisionFrame:
 			'input': temp_vision_frame
 		})[0]
 
-		if model_name == 'u2net_cloth':
+		if model_type == 'u2net_cloth':
 			remove_vision_frame = numpy.argmax(remove_vision_frame, axis = 1)
 
 	return remove_vision_frame
 
 
+def forward_corridor_key(temp_vision_frame : VisionFrame) -> Tuple[Mask, VisionFrame]:
+	background_remover = get_inference_pool().get('background_remover')
+
+	with thread_semaphore():
+		remove_vision_mask, remove_vision_frame = background_remover.run(None,
+		{
+			'input': temp_vision_frame
+		})
+
+	return remove_vision_mask, remove_vision_frame
+
+
 def prepare_temp_frame(temp_vision_frame : VisionFrame) -> VisionFrame:
+	model_type = get_model_options().get('type')
 	model_size = get_model_options().get('size')
 	model_mean = get_model_options().get('mean')
 	model_standard_deviation = get_model_options().get('standard_deviation')
 
+	if model_type == 'corridor_key':
+		coarse_color = temp_vision_frame[:, :, ::-1].astype(numpy.float32) / 255.0
+		coarse_bias = coarse_color[:, :, 1] - numpy.maximum(coarse_color[:, :, 0], coarse_color[:, :, 2])
+		coarse_vision_mask = cv2.resize(1.0 - numpy.clip(coarse_bias * 2.0, 0, 1), model_size)[:, :, numpy.newaxis]
+
 	temp_vision_frame = cv2.resize(temp_vision_frame, model_size)
 	temp_vision_frame = temp_vision_frame[:, :, ::-1] / 255.0
 	temp_vision_frame = (temp_vision_frame - model_mean) / model_standard_deviation
+
+	if model_type == 'corridor_key':
+		temp_vision_frame = numpy.concatenate([ temp_vision_frame, coarse_vision_mask ], axis = 2)
+
 	temp_vision_frame = temp_vision_frame.transpose(2, 0, 1)
 	temp_vision_frame = numpy.expand_dims(temp_vision_frame, axis = 0).astype(numpy.float32)
 	return temp_vision_frame
@@ -516,16 +629,32 @@ def normalize_vision_mask(temp_vision_mask : Mask) -> Mask:
 	return temp_vision_mask
 
 
-def apply_background_color(temp_vision_frame : VisionFrame, temp_vision_mask : Mask) -> VisionFrame:
-	background_remover_color = state_manager.get_item('background_remover_color')
+def apply_fill_color(temp_vision_frame : VisionFrame, temp_vision_mask : Mask) -> VisionFrame:
+	background_remover_fill_color = state_manager.get_item('background_remover_fill_color')
 	temp_vision_mask = temp_vision_mask.astype(numpy.float32) / 255
 	temp_vision_mask = numpy.expand_dims(temp_vision_mask, axis = 2)
-	temp_vision_mask = (1 - temp_vision_mask) * background_remover_color[-1] / 255
-	color_frame = numpy.zeros_like(temp_vision_frame)
-	color_frame[:, :, 0] = background_remover_color[2]
-	color_frame[:, :, 1] = background_remover_color[1]
-	color_frame[:, :, 2] = background_remover_color[0]
-	temp_vision_frame = temp_vision_frame * (1 - temp_vision_mask) + color_frame * temp_vision_mask
+	temp_vision_mask = (1 - temp_vision_mask) * background_remover_fill_color[-1] / 255
+	fill_vision_frame = numpy.zeros_like(temp_vision_frame)
+	fill_vision_frame[:, :, 0] = background_remover_fill_color[2]
+	fill_vision_frame[:, :, 1] = background_remover_fill_color[1]
+	fill_vision_frame[:, :, 2] = background_remover_fill_color[0]
+	temp_vision_frame = temp_vision_frame * (1 - temp_vision_mask) + fill_vision_frame * temp_vision_mask
+	temp_vision_frame = temp_vision_frame.astype(numpy.uint8)
+	return temp_vision_frame
+
+
+def apply_despill_color(temp_vision_frame : VisionFrame) -> VisionFrame:
+	background_remover_despill_color = state_manager.get_item('background_remover_despill_color')
+	temp_vision_frame = temp_vision_frame.astype(numpy.float32)
+	color_alpha = background_remover_despill_color[3] / 255.0
+	despill_vision_frame = numpy.zeros_like(temp_vision_frame)
+	despill_vision_frame[:, :, 0] = background_remover_despill_color[2]
+	despill_vision_frame[:, :, 1] = background_remover_despill_color[1]
+	despill_vision_frame[:, :, 2] = background_remover_despill_color[0]
+	color_weight = despill_vision_frame / numpy.maximum(numpy.max(background_remover_despill_color[:3]), 1)
+	color_limit = numpy.roll(temp_vision_frame, 1, 2) + numpy.roll(temp_vision_frame, -1, 2)
+	limit_vision_frame = numpy.minimum(temp_vision_frame, color_limit * 0.5)
+	temp_vision_frame = temp_vision_frame + (limit_vision_frame - temp_vision_frame) * color_alpha * color_weight
 	temp_vision_frame = temp_vision_frame.astype(numpy.uint8)
 	return temp_vision_frame