diaojiaolou
/
Deep-Live-Cam
mirror of https://github.com/hacksider/Deep-Live-Cam.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

feat(): adding mouth masking

pull/738/head
GhoulBoii 2024-10-24 00:24:42 +05:30
parent 6f6f93a4ad
commit d616c513c9
No known key found for this signature in database
GPG Key ID: F443BDA92506073F
3 changed files with 208 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
modules/globals.py
View File

@ -36,3 +36,6 @@ fp_ui: Dict[str, bool] = {"face_enhancer": False}
camera_input_combobox = None
webcam_preview_running = False
show_fps = False
mouth_mask = False
show_mouth_mask_box = False
mouth_mask_switch_preview = None

161
modules/processors/frame/face_swapper.py
View File

@ -1,5 +1,6 @@
from typing import Any, List
from typing import Any, List, Tuple
import cv2
import numpy as np
import insightface
import threading
@ -8,29 +9,48 @@ import modules.processors.frame.core
from modules.core import update_status
from modules.face_analyser import get_one_face, get_many_faces, default_source_face
from modules.typing import Face, Frame
from modules.utilities import conditional_download, resolve_relative_path, is_image, is_video
from modules.utilities import (
conditional_download,
resolve_relative_path,
is_image,
is_video,
)
from modules.cluster_analysis import find_closest_centroid
FACE_SWAPPER = None
THREAD_LOCK = threading.Lock()
NAME = 'DLC.FACE-SWAPPER'
NAME = "DLC.FACE-SWAPPER"
# Add mouth landmarks indices for masking
MOUTH_LANDMARKS = list(
range(46, 68)
) # Common indices for mouth landmarks in facial detection
def pre_check() -> bool:
download_directory_path = resolve_relative_path('../models')
conditional_download(download_directory_path, ['https://huggingface.co/hacksider/deep-live-cam/blob/main/inswapper_128_fp16.onnx'])
download_directory_path = resolve_relative_path("../models")
conditional_download(
download_directory_path,
[
"https://huggingface.co/hacksider/deep-live-cam/blob/main/inswapper_128_fp16.onnx"
],
)
return True
def pre_start() -> bool:
if not modules.globals.map_faces and not is_image(modules.globals.source_path):
update_status('Select an image for source path.', NAME)
update_status("Select an image for source path.", NAME)
return False
elif not modules.globals.map_faces and not get_one_face(cv2.imread(modules.globals.source_path)):
update_status('No face in source path detected.', NAME)
elif not modules.globals.map_faces and not get_one_face(
cv2.imread(modules.globals.source_path)
):
update_status("No face in source path detected.", NAME)
return False
if not is_image(modules.globals.target_path) and not is_video(modules.globals.target_path):
update_status('Select an image or video for target path.', NAME)
if not is_image(modules.globals.target_path) and not is_video(
modules.globals.target_path
):
update_status("Select an image or video for target path.", NAME)
return False
return True
@ -40,20 +60,60 @@ def get_face_swapper() -> Any:
with THREAD_LOCK:
if FACE_SWAPPER is None:
model_path = resolve_relative_path('../models/inswapper_128_fp16.onnx')
FACE_SWAPPER = insightface.model_zoo.get_model(model_path, providers=modules.globals.execution_providers)
model_path = resolve_relative_path("../models/inswapper_128_fp16.onnx")
FACE_SWAPPER = insightface.model_zoo.get_model(
model_path, providers=modules.globals.execution_providers
)
return FACE_SWAPPER
def create_mouth_mask(face: Face, frame_shape: Tuple[int, int]) -> np.ndarray:
"""Create a mask for the mouth region"""
mask = np.zeros(frame_shape[:2], dtype=np.uint8)
# Get mouth landmarks from the face
landmarks = face.kps
mouth_points = landmarks[MOUTH_LANDMARKS].astype(np.int32)
# Create a polygon around the mouth region
cv2.fillPoly(mask, [mouth_points], 255)
# Dilate the mask slightly to ensure smooth blending
kernel = np.ones((5, 5), np.uint8)
mask = cv2.dilate(mask, kernel, iterations=2)
# Blur the mask edges
mask = cv2.GaussianBlur(mask, (15, 15), 10)
return mask
def blend_with_mask(
swapped_frame: Frame, original_frame: Frame, mask: np.ndarray
) -> Frame:
"""Blend the swapped face with the original frame using the mouth mask"""
mask_3channel = cv2.cvtColor(mask, cv2.COLOR_GRAY2BGR) / 255.0
# Blend the images based on the mask
blended = swapped_frame * (1 - mask_3channel) + original_frame * mask_3channel
return blended.astype(np.uint8)
def swap_face(source_face: Face, target_face: Face, temp_frame: Frame) -> Frame:
return get_face_swapper().get(temp_frame, target_face, source_face, paste_back=True)
# Store the original frame for mouth preservation
original_frame = temp_frame.copy()
# Perform the face swap
swapped_frame = get_face_swapper().get(
temp_frame, target_face, source_face, paste_back=True
)
def process_frame(source_face: Face, temp_frame: Frame) -> Frame:
# Ensure the frame is in RGB format if color correction is enabled
if modules.globals.color_correction:
temp_frame = cv2.cvtColor(temp_frame, cv2.COLOR_BGR2RGB)
if modules.globals.many_faces:
many_faces = get_many_faces(temp_frame)
if many_faces:
@ -71,34 +131,42 @@ def process_frame_v2(temp_frame: Frame, temp_frame_path: str = "") -> Frame:
if modules.globals.many_faces:
source_face = default_source_face()
for map in modules.globals.souce_target_map:
target_face = map['target']['face']
target_face = map["target"]["face"]
temp_frame = swap_face(source_face, target_face, temp_frame)
elif not modules.globals.many_faces:
for map in modules.globals.souce_target_map:
if "source" in map:
source_face = map['source']['face']
target_face = map['target']['face']
source_face = map["source"]["face"]
target_face = map["target"]["face"]
temp_frame = swap_face(source_face, target_face, temp_frame)
elif is_video(modules.globals.target_path):
if modules.globals.many_faces:
source_face = default_source_face()
for map in modules.globals.souce_target_map:
target_frame = [f for f in map['target_faces_in_frame'] if f['location'] == temp_frame_path]
target_frame = [
f
for f in map["target_faces_in_frame"]
if f["location"] == temp_frame_path
]
for frame in target_frame:
for target_face in frame['faces']:
for target_face in frame["faces"]:
temp_frame = swap_face(source_face, target_face, temp_frame)
elif not modules.globals.many_faces:
for map in modules.globals.souce_target_map:
if "source" in map:
target_frame = [f for f in map['target_faces_in_frame'] if f['location'] == temp_frame_path]
source_face = map['source']['face']
target_frame = [
f
for f in map["target_faces_in_frame"]
if f["location"] == temp_frame_path
]
source_face = map["source"]["face"]
for frame in target_frame:
for target_face in frame['faces']:
for target_face in frame["faces"]:
temp_frame = swap_face(source_face, target_face, temp_frame)
else:
detected_faces = get_many_faces(temp_frame)
@ -110,25 +178,46 @@ def process_frame_v2(temp_frame: Frame, temp_frame_path: str = "") -> Frame:
elif not modules.globals.many_faces:
if detected_faces:
if len(detected_faces) <= len(modules.globals.simple_map['target_embeddings']):
if len(detected_faces) <= len(
modules.globals.simple_map["target_embeddings"]
):
for detected_face in detected_faces:
closest_centroid_index, _ = find_closest_centroid(modules.globals.simple_map['target_embeddings'], detected_face.normed_embedding)
closest_centroid_index, _ = find_closest_centroid(
modules.globals.simple_map["target_embeddings"],
detected_face.normed_embedding,
)
temp_frame = swap_face(modules.globals.simple_map['source_faces'][closest_centroid_index], detected_face, temp_frame)
temp_frame = swap_face(
modules.globals.simple_map["source_faces"][
closest_centroid_index
],
detected_face,
temp_frame,
)
else:
detected_faces_centroids = []
for face in detected_faces:
detected_faces_centroids.append(face.normed_embedding)
detected_faces_centroids.append(face.normed_embedding)
i = 0
for target_embedding in modules.globals.simple_map['target_embeddings']:
closest_centroid_index, _ = find_closest_centroid(detected_faces_centroids, target_embedding)
for target_embedding in modules.globals.simple_map[
"target_embeddings"
]:
closest_centroid_index, _ = find_closest_centroid(
detected_faces_centroids, target_embedding
)
temp_frame = swap_face(modules.globals.simple_map['source_faces'][i], detected_faces[closest_centroid_index], temp_frame)
temp_frame = swap_face(
modules.globals.simple_map["source_faces"][i],
detected_faces[closest_centroid_index],
temp_frame,
)
i += 1
return temp_frame
def process_frames(source_path: str, temp_frame_paths: List[str], progress: Any = None) -> None:
def process_frames(
source_path: str, temp_frame_paths: List[str], progress: Any = None
) -> None:
if not modules.globals.map_faces:
source_face = get_one_face(cv2.imread(source_path))
for temp_frame_path in temp_frame_paths:
@ -162,7 +251,9 @@ def process_image(source_path: str, target_path: str, output_path: str) -> None:
cv2.imwrite(output_path, result)
else:
if modules.globals.many_faces:
update_status('Many faces enabled. Using first source image. Progressing...', NAME)
update_status(
"Many faces enabled. Using first source image. Progressing...", NAME
)
target_frame = cv2.imread(output_path)
result = process_frame_v2(target_frame)
cv2.imwrite(output_path, result)
@ -170,5 +261,9 @@ def process_image(source_path: str, target_path: str, output_path: str) -> None:
def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
if modules.globals.map_faces and modules.globals.many_faces:
update_status('Many faces enabled. Using first source image. Progressing...', NAME)
modules.processors.frame.core.process_video(source_path, temp_frame_paths, process_frames)
update_status(
"Many faces enabled. Using first source image. Progressing...", NAME
)
modules.processors.frame.core.process_video(
source_path, temp_frame_paths, process_frames
)

77
modules/ui.py
View File

@ -59,6 +59,8 @@ RECENT_DIRECTORY_SOURCE = None
RECENT_DIRECTORY_TARGET = None
RECENT_DIRECTORY_OUTPUT = None
BLUR_SIZE = 1
preview_label = None
preview_slider = None
source_label = None
@ -317,6 +319,47 @@ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.C
live_button.place(relx=0.65, rely=0.86, relwidth=0.2, relheight=0.05)
# --- End Camera Selection ---
show_mouth_mask_var = ctk.BooleanVar(value=modules.globals.show_mouth_mask_box)
show_mouth_mask_switch = ctk.CTkSwitch(
root,
text="Show Mouth Mask Box",
variable=show_mouth_mask_var,
cursor="hand2",
command=lambda: setattr(
modules.globals, "show_mouth_mask_box", show_mouth_mask_var.get()
),
progress_color="#3a7ebf",
font=("Roboto", 14, "bold"),
)
show_mouth_mask_switch.pack(pady=5, anchor="w")
# Create frame for mouth mask
mouth_mask_frame = ctk.CTkFrame(
root, fg_color="#2a2d2e", border_width=1, border_color="grey"
)
mouth_mask_frame.pack(pady=5, anchor="w", fill="x")
# Create a shared BooleanVar in modules.globals
if not hasattr(modules.globals, "mouth_mask_var"):
modules.globals.mouth_mask_var = ctk.BooleanVar(
value=modules.globals.mouth_mask
)
# Mouth mask switch
mouth_mask_switch = ctk.CTkSwitch(
mouth_mask_frame,
text="Mouth Mask",
variable=modules.globals.mouth_mask_var,
cursor="hand2",
command=toggle_mouthmask,
progress_color="#3a7ebf",
font=("Roboto", 14, "bold"),
)
mouth_mask_switch.pack(pady=5, anchor="w")
# Store the switch in modules.globals for access from create_preview
modules.globals.mouth_mask_switch_root = mouth_mask_switch
status_label = ctk.CTkLabel(root, text=None, justify="center")
status_label.place(relx=0.1, rely=0.9, relwidth=0.8)
@ -478,6 +521,29 @@ def update_popup_source(
return map
def toggle_mouthmask():
"""
Toggle the mouth mask state and synchronize all UI switches.
Updates both the global state and any existing switch controls.
"""
is_mouthmask = modules.globals.mouth_mask_var.get()
modules.globals.mouth_mask = is_mouthmask
# Update root window switch if it exists
if hasattr(modules.globals, "mouth_mask_switch_root"):
if is_mouthmask:
modules.globals.mouth_mask_switch_root.select()
else:
modules.globals.mouth_mask_switch_root.deselect()
# Update preview window switch if it exists
if hasattr(modules.globals, "mouth_mask_switch_preview"):
if is_mouthmask:
modules.globals.mouth_mask_switch_preview.select()
else:
modules.globals.mouth_mask_switch_preview.deselect()
def create_preview(parent: ctk.CTkToplevel) -> ctk.CTkToplevel:
global preview_label, preview_slider
@ -494,6 +560,17 @@ def create_preview(parent: ctk.CTkToplevel) -> ctk.CTkToplevel:
preview_slider = ctk.CTkSlider(
preview, from_=0, to=0, command=lambda frame_value: update_preview(frame_value)
)
mouth_mask_switch_preview = ctk.CTkSwitch(
preview,
text="Mouth Mask",
variable=modules.globals.mouth_mask_var,
cursor="hand2",
command=toggle_mouthmask,
)
mouth_mask_switch_preview.pack(side="left", padx=5, pady=5)
# Store the switch in modules.globals for access from create_root
modules.globals.mouth_mask_switch_preview = mouth_mask_switch_preview
return preview