deal with mask

app.py

@@ -1,6 +1,8 @@
+import cv2
+import numpy as np
+from PIL import Image
 import os
 import gradio as gr
-from scepter.modules.utils.file_system import FS
 from huggingface_hub import hf_hub_download, snapshot_download
 
 def resolve_hf_path(path):
@@ -40,12 +42,44 @@ from modules.flux import FluxMRModiACEPlus
 from inference.registry import INFERENCES
 
 
-config_path = os.path.join("config", "simple_faceswap.yaml")
+config_path = os.path.join("config", "ace_plus_fft.yaml")
 cfg = Config(load=True, cfg_file=config_path)
 
 # Instantiate the ACEInference object.
 ace_infer = ACEInference(cfg)
 
+def create_face_mask(pil_image):
+    """
+    Create a binary mask (PIL Image) from a PIL image by detecting the face region.
+    The mask will be white (255) on the detected face area and black (0) elsewhere.
+    """
+    # Convert PIL image to a numpy array in RGB format
+    image_np = np.array(pil_image.convert("RGB"))
+    # Convert to grayscale for face detection
+    gray = cv2.cvtColor(image_np, cv2.COLOR_RGB2GRAY)
+
+    # Load the Haar cascade for face detection (make sure opencv data is installed)
+    cascade_path = cv2.data.haarcascades + "haarcascade_frontalface_default.xml"
+    face_cascade = cv2.CascadeClassifier(cascade_path)
+
+    # Detect faces in the image
+    faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5)
+
+    # Create an empty mask with the same dimensions as the image
+    mask = np.zeros_like(gray, dtype=np.uint8)
+
+    # For each detected face, draw a white rectangle (or a more refined shape)
+    for (x, y, w, h) in faces:
+        # Optionally expand the bounding box slightly
+        padding = 0.2
+        x1 = max(0, int(x - w * padding))
+        y1 = max(0, int(y - h * padding))
+        x2 = min(gray.shape[1], int(x + w * (1 + padding)))
+        y2 = min(gray.shape[0], int(y + h * (1 + padding)))
+        mask[y1:y2, x1:x2] = 255
+
+    return Image.fromarray(mask)
+
 def face_swap_app(target_img, face_img):
     if target_img is None or face_img is None:
         raise ValueError("Both a target image and a face image must be provided.")
@@ -57,14 +91,14 @@ def face_swap_app(target_img, face_img):
     output_img, edit_image, change_image, mask, seed = ace_infer(
         reference_image=target_img,
         edit_image=face_img,
-        edit_mask=None,          # Let ACE++ generate the mask automatically
+        edit_mask=None,
         prompt="Face swap",
         output_height=1024,
         output_width=1024,
         sampler='flow_euler',
         sample_steps=28,
         guide_scale=50,
-        seed=-1                  # Random seed if not provided
+        seed=-1
     )
     return output_img
 

config/simple_faceswap.yaml

@@ -1,17 +0,0 @@
-NAME: SimpleFaceSwapConfig
-MODEL:
-  LOCAL_EDITING:
-    MODEL_PATH: ${LOCAL_MODEL_PATH}
-    REPAINTING_SCALE: 0.5
-    PREPROCESSOR:
-      - NAME: CannyAnnotator
-        TYPE: canny_repainting
-        LOW_THRESHOLD: 100
-        HIGH_THRESHOLD: 200
-      - NAME: ColorAnnotator
-        TYPE: mosaic_repainting
-        RATIO: 64
-DTYPE: bfloat16
-MAX_SEQ_LEN: 77
-SAMPLE_ARGS:
-  prompt: "Face swap"

requirements.txt

@@ -4,4 +4,7 @@ scepter
 torch
 torchvision
 transformers
-huggingface_hub
+huggingface_hub
+Pillow
+numpy
+opencv-python