MediapipeFacialSegment node support face_landmarker to compatible Python 3.12 and 3.13

py/mediapipe_facial_segment.py

@@ -1,41 +1,65 @@
-# layerstyle advance
+import os
-
 import numpy as np
 from .imagefunc import *
 
 NODE_NAME = 'MediapipeFacialSegment'
 
 
-# 获取特征点的坐标
 def get_points(indices, face_landmarks, width, height):
-    return [(int(face_landmarks.landmark[i].x * width), int(face_landmarks.landmark[i].y * height))
+    return [
-            for i in indices]
+        (
+            int(face_landmarks.landmark[i].x * width),
+            int(face_landmarks.landmark[i].y * height)
+        )
+        for i in indices
+    ]
+
 
-# 绘制面部特征的多边形
 def draw_feature(indices, mask, face_landmarks, width, height):
     points = get_points(indices, face_landmarks, width, height)
     points = np.array(points, dtype=np.int32)
     cv2.fillPoly(mask, [points], 255)
 
+
+class _Landmark:
+    __slots__ = ("x", "y")
+
+    def __init__(self, x, y):
+        self.x = x
+        self.y = y
+
+
+class _FaceLandmarksWrapper:
+    def __init__(self, landmarks):
+        self.landmark = [_Landmark(lm.x, lm.y) for lm in landmarks]
+
+
+def _get_face_landmarker_model_path():
+    base_dir = os.path.dirname(__file__)           # .../ComfyUI_LayerStyle_Advance/py
+    plugin_dir = os.path.dirname(base_dir)         # .../ComfyUI_LayerStyle_Advance
+    return os.path.join(
+        plugin_dir,
+        "face_landmarker",
+        "face_landmarker.task"
+    )
+
+
 class FacialFeatureSegment:
 
     def __init__(self):
-        pass
+        self._tasks_landmarker = None
 
     @classmethod
     def INPUT_TYPES(self):
-
         return {
             "required": {
-                "image": ("IMAGE",),  #
+                "image": ("IMAGE",),
                 "left_eye": ("BOOLEAN", {"default": True}),
                 "left_eyebrow": ("BOOLEAN", {"default": True}),
                 "right_eye": ("BOOLEAN", {"default": True}),
                 "right_eyebrow": ("BOOLEAN", {"default": True}),
                 "lips": ("BOOLEAN", {"default": True}),
                 "tooth": ("BOOLEAN", {"default": True}),
-            },
-            "optional": {
             }
         }
 
@@ -44,42 +68,86 @@ class FacialFeatureSegment:
     FUNCTION = 'facial_feature_segment'
     CATEGORY = '😺dzNodes/LayerMask'
 
-    def facial_feature_segment(self, image,
+    def facial_feature_segment(
+        self,
+        image,
         left_eye, left_eyebrow, right_eye, right_eyebrow, lips, tooth
     ):
-
         import mediapipe as mp
-        # 定义面部特征索引
+
+        USE_SOLUTIONS = hasattr(mp, "solutions")
+
+
         left_eye_indices = [33, 7, 163, 144, 145, 153, 154, 155, 133, 173, 157, 158, 159, 160, 161, 246]
         right_eye_indices = [263, 249, 390, 373, 374, 380, 381, 382, 362, 398, 384, 385, 386, 387, 388, 466]
         left_eyebrow_indices = [70, 63, 105, 66, 107, 55, 65, 52, 53, 46]
         right_eyebrow_indices = [336, 296, 334, 293, 300, 276, 283, 282, 295, 285]
-        # upper_lip_indices = [61, 146, 91, 181, 84, 17, 314, 405, 321, 375, 291, 308, 324, 318, 402, 317, 14, 87, 178, 88, 95, 78]
-        # lower_lip_indices = [61, 185, 40, 39, 37, 0, 267, 269, 270, 409, 291, 308, 415, 310, 311, 312, 13, 82, 81, 80, 191, 78]
         tooth_indices = [78, 95, 88, 178, 87, 14, 317, 402, 318, 324, 308, 415, 310, 311, 312, 13, 82, 81, 80, 191, 78]
-        lips_indices = [61, 76, 62, 78, 191, 80, 81, 82, 13, 312, 311, 310, 415, 308, 324, 318, 402, 317, 14, 87, 178,
+        lips_indices = [
-                         88, 95, 185, 40, 39, 37, 0, 267, 269, 270, 409, 291, 375, 321, 405, 314, 17, 84, 181, 91, 146,
+            61, 76, 62, 78, 191, 80, 81, 82, 13, 312, 311, 310, 415, 308,
-                         61]
+            324, 318, 402, 317, 14, 87, 178, 88, 95, 185, 40, 39, 37, 0,
+            267, 269, 270, 409, 291, 375, 321, 405, 314, 17, 84, 181, 91,
+            146, 61
+        ]
 
         ret_images = []
         ret_masks = []
         scale_factor = 4
 
+
+        if USE_SOLUTIONS:
+            mp_face_mesh = mp.solutions.face_mesh
+            face_mesh = mp_face_mesh.FaceMesh(
+                static_image_mode=True,
+                max_num_faces=1,
+                min_detection_confidence=0.5
+            )
+        else:
+            from mediapipe.tasks.python import vision
+            from mediapipe.tasks.python import BaseOptions
+
+            if self._tasks_landmarker is None:
+                model_path = _get_face_landmarker_model_path()
+
+                if not os.path.exists(model_path):
+                    raise RuntimeError(
+                        f"[{NODE_NAME}] FaceLandmarker model not found:\n{model_path}"
+                    )
+
+                options = vision.FaceLandmarkerOptions(
+                    base_options=BaseOptions(model_asset_path=model_path),
+                    output_face_blendshapes=False,
+                    output_facial_transformation_matrixes=False,
+                    num_faces=1
+                )
+
+                self._tasks_landmarker = vision.FaceLandmarker.create_from_options(options)
+
+
         for i in image:
             face_image = tensor2pil(i.unsqueeze(0)).convert('RGB')
             width, height = face_image.size
             width *= scale_factor
             height *= scale_factor
+
             cv2_image = pil2cv2(face_image)
-            mp_face_mesh = mp.solutions.face_mesh
-            fase_mesh = mp_face_mesh.FaceMesh(static_image_mode=True, max_num_faces=1, min_detection_confidence=0.5)
-            results = fase_mesh.process(cv2_image)
             mask = np.zeros((height, width), dtype=np.uint8)
 
+            if USE_SOLUTIONS:
+                results = face_mesh.process(cv2_image)
+                faces = results.multi_face_landmarks or []
+            else:
+                mp_image = mp.Image(
+                    image_format=mp.ImageFormat.SRGB,
+                    data=cv2_image
+                )
+                result = self._tasks_landmarker.detect(mp_image)
+                faces = [
+                    _FaceLandmarksWrapper(lms)
+                    for lms in (result.face_landmarks or [])
+                ]
 
-            if results.multi_face_landmarks:
+            for face_landmarks in faces:
-                for face_landmarks in results.multi_face_landmarks:
-                    # 绘制各个面部特征
                 if left_eye:
                     draw_feature(left_eye_indices, mask, face_landmarks, width, height)
                 if right_eye:
@@ -96,12 +164,14 @@ class FacialFeatureSegment:
             mask = cv22pil(mask).convert('L')
             mask = gaussian_blur(mask, 2)
             mask = mask.resize(face_image.size, Image.BILINEAR)
+
             ret_images.append(pil2tensor(RGB2RGBA(face_image, mask)))
             ret_masks.append(image2mask(mask))
 
         log(f"{NODE_NAME} Processed {len(ret_images)} image(s).", message_type='finish')
         return (torch.cat(ret_images, dim=0), torch.cat(ret_masks, dim=0),)
 
+
 NODE_CLASS_MAPPINGS = {
     "LayerMask: MediapipeFacialSegment": FacialFeatureSegment
 }

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "ComfyUI_LayerStyle_Advance"
 description = "The nodes detached from ComfyUI Layer Style are mainly those with complex requirements for dependency packages."
-version = "2.0.34"
+version = "2.0.35"
 license = { text = "MIT License" }
 dependencies = ["numpy", "matplotlib", "scikit_image", "scikit_learn", "opencv-contrib-python", "pymatting", "timm", "blend_modes", "transformers", "diffusers", "loguru", "colour-science", "huggingface_hub", "segment_anything", "addict", "omegaconf", "yapf", "wget", "iopath", "mediapipe", "typer_config", "fastapi", "rich", "google-generativeai", "ultralytics", "transparent-background", "accelerate", "onnxruntime", "bitsandbytes", "peft", "protobuf", "hydra-core", "blind-watermark", "qrcode", "pyzbar", "psd-tools", "wandb", "zhipuai", "openai","google-genai", "fastapi","typer-config"]