Merge pull request #6 from ArtVentureX/main

Fix ModuleNotFoundError and improve start up time

__init__.py

@@ -8,6 +8,7 @@
 from .log import log, blue_text, cyan_text, get_summary, get_label
 
 #Ref: https://github.com/comfyanonymous/ComfyUI/blob/76d53c4622fc06372975ed2a43ad345935b8a551/nodes.py#L17
+sys.path.insert(0, str(Path(here, "src").resolve()))
 for pkg_name in os.listdir(str(Path(here, "src"))):
     sys.path.insert(0, str(Path(here, "src", pkg_name).resolve()))
 print(f"Registered sys.path: {sys.path}")

node_wrappers/binary.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.binary import BinaryDetector
 import comfy.model_management as model_management
 
 class Binary_Preprocessor:
@@ -18,6 +17,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, bin_threshold, **kwargs):
+        from controlnet_aux.binary import BinaryDetector
+
         return (common_annotator_call(BinaryDetector(), image, bin_threshold=bin_threshold), )
 
 

node_wrappers/canny.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.canny import CannyDetector
 import comfy.model_management as model_management
 
 class Canny_Edge_Preprocessor:
@@ -19,6 +18,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, low_threshold, high_threshold, **kwargs):
+        from controlnet_aux.canny import CannyDetector
+
         return (common_annotator_call(CannyDetector(), image, low_threshold=low_threshold, high_threshold=high_threshold), )
 
 

node_wrappers/color.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.color import ColorDetector
 import comfy.model_management as model_management
 
 class Color_Preprocessor:
@@ -15,6 +14,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/T2IAdapter-only"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.color import ColorDetector
+
         return (common_annotator_call(ColorDetector(), image), )
 
 

node_wrappers/dwpose.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME, DWPOSE_MODEL_NAME
-from controlnet_aux.dwpose import DwposeDetector
 import comfy.model_management as model_management
 
 class DWPose_Preprocessor:
@@ -19,6 +18,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Faces and Poses"
 
     def estimate_pose(self, image, detect_hand, detect_body, detect_face, **kwargs):
+        from controlnet_aux.dwpose import DwposeDetector
+
         detect_hand = detect_hand == "enable"
         detect_body = detect_body == "enable"
         detect_face = detect_face == "enable"

node_wrappers/hed.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.hed import HEDdetector
 import comfy.model_management as model_management
 
 class HED_Preprocessor:
@@ -18,6 +17,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.hed import HEDdetector
+
         model = HEDdetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, safe = kwargs["safe"] == "enable")
         del model

node_wrappers/leres.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.leres import LeresDetector
 import comfy.model_management as model_management
 
 class LERES_Depth_Map_Preprocessor:
@@ -22,6 +21,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Normal and Depth Map"
 
     def execute(self, image, rm_nearest, rm_background, **kwargs):
+        from controlnet_aux.leres import LeresDetector
+
         model = LeresDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, thr_a=rm_nearest, thr_b=rm_background, boost=kwargs["boost"] == "enable")
         del model

node_wrappers/lineart.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.lineart import LineartDetector
 import comfy.model_management as model_management
 
 class LineArt_Preprocessor:
@@ -18,6 +17,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.lineart import LineartDetector
+
         model = LineartDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, coarse = kwargs["coarse"] == "enable")
         del model

node_wrappers/lineart_anime.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.lineart_anime import LineartAnimeDetector
 import comfy.model_management as model_management
 
 class AnimeLineArt_Preprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.lineart_anime import LineartAnimeDetector
+
         model = LineartAnimeDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
         del model

node_wrappers/manga_line.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.manga_line import LineartMangaDetector
 import comfy.model_management as model_management
 
 class Manga2Anime_LineArt_Preprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.manga_line import LineartMangaDetector
+
         model = LineartMangaDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
         del model

node_wrappers/mediapipe_face.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME, DWPOSE_MODEL_NAME
-from controlnet_aux.mediapipe_face import MediapipeFaceDetector
 import comfy.model_management as model_management
 
 class Media_Pipe_Face_Mesh_Preprocessor:
@@ -15,6 +14,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Faces and Poses"
 
     def detect(self, image, max_faces, min_confidence):
+        from controlnet_aux.mediapipe_face import MediapipeFaceDetector
+
         return (common_annotator_call(MediapipeFaceDetector(), image, max_faces=max_faces, min_confidence=min_confidence), )
 
 NODE_CLASS_MAPPINGS = {

node_wrappers/midas.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.midas import MidasDetector
 import comfy.model_management as model_management
 import numpy as np
 
@@ -17,6 +16,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Normal and Depth Map"
 
     def execute(self, image, a, bg_threshold, **kwargs):
+        from controlnet_aux.midas import MidasDetector
+
         model = MidasDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         #Dirty hack :))
         cb = lambda image, **kargs: model(image, **kargs)[1]
@@ -38,6 +39,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Normal and Depth Map"
 
     def execute(self, image, a, bg_threshold, **kwargs):
+        from controlnet_aux.midas import MidasDetector
+
         # Ref: https://github.com/lllyasviel/ControlNet/blob/main/gradio_depth2image.py
         model = MidasDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, a=a, bg_th=bg_threshold)

node_wrappers/mlsd.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.mlsd import MLSDdetector
 import comfy.model_management as model_management
 import numpy as np
 
@@ -20,6 +19,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, score_threshold, dist_threshold, **kwargs):
+        from controlnet_aux.mlsd import MLSDdetector
+
         model = MLSDdetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, thr_v=score_threshold, thr_d=dist_threshold)
         return (out, )

node_wrappers/normalbae.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.normalbae import NormalBaeDetector
 import comfy.model_management as model_management
 
 class BAE_Normal_Map_Preprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Normal and Depth Map"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.normalbae import NormalBaeDetector
+
         model = NormalBaeDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
         del model

node_wrappers/oneformer.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.oneformer import OneformerSegmentor
 import comfy.model_management as model_management
 
 class OneFormer_COCO_SemSegPreprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Semantic Segmentation"
 
     def semantic_segmentate(self, image):
+        from controlnet_aux.oneformer import OneformerSegmentor
+
         model = OneformerSegmentor.from_pretrained(HF_MODEL_NAME, "150_16_swin_l_oneformer_coco_100ep.pth", cache_dir=annotator_ckpts_path)
         model = model.to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
@@ -30,6 +31,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Semantic Segmentation"
 
     def semantic_segmentate(self, image):
+        from controlnet_aux.oneformer import OneformerSegmentor
+
         model = OneformerSegmentor.from_pretrained(HF_MODEL_NAME, "250_16_swin_l_oneformer_ade20k_160k.pth", cache_dir=annotator_ckpts_path)
         model = model.to(model_management.get_torch_device())
         out = common_annotator_call(model, image)

node_wrappers/openpose.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME, DWPOSE_MODEL_NAME
-from controlnet_aux.open_pose import OpenposeDetector
 import comfy.model_management as model_management
 
 class OpenPose_Preprocessor:
@@ -19,6 +18,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors"
 
     def estimate_pose(self, image, detect_hand, detect_body, detect_face, **kwargs):
+        from controlnet_aux.open_pose import OpenposeDetector
+
         detect_hand = detect_hand == "enable"
         detect_body = detect_body == "enable"
         detect_face = detect_face == "enable"

node_wrappers/pidinet.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.pidi import PidiNetDetector
 import comfy.model_management as model_management
 
 class PIDINET_Preprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, safe, **kwargs):
+        from controlnet_aux.pidi import PidiNetDetector
+
         model = PidiNetDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image, safe = safe == "enable")
         del model

node_wrappers/scribble.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.scribble import ScribbleDetector, ScribbleXDog_Detector
 import comfy.model_management as model_management
 
 class Scribble_Preprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.scribble import ScribbleDetector
+
         model = ScribbleDetector()
         return (common_annotator_call(model, image), )
 
@@ -32,6 +33,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Line Extractors"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.scribble import ScribbleXDog_Detector
+
         model = ScribbleXDog_Detector()
         return (common_annotator_call(model, image), )
 

node_wrappers/segment_anything.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.segment_anything import SamDetector
 import comfy.model_management as model_management
 
 class SAM_Preprocessor:
@@ -15,6 +14,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/others"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.segment_anything import SamDetector
+
         mobile_sam = SamDetector.from_pretrained("dhkim2810/MobileSAM", model_type="vit_t", filename="mobile_sam.pt").to(model_management.get_torch_device())
         out = common_annotator_call(mobile_sam, image)
         del mobile_sam

node_wrappers/shuffle.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.shuffle import ContentShuffleDetector
 import comfy.model_management as model_management
 
 class Shuffle_Preprocessor:
@@ -12,6 +11,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/T2IAdapter-only"
 
     def preprocess(self, image):
+        from controlnet_aux.shuffle import ContentShuffleDetector
+
         return (common_annotator_call(ContentShuffleDetector(), image), )
 
 NODE_CLASS_MAPPINGS = {

node_wrappers/uniformer.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.uniformer import UniformerSegmentor
 import comfy.model_management as model_management
 
 class Uniformer_SemSegPreprocessor:
@@ -13,6 +12,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Semantic Segmentation"
 
     def semantic_segmentate(self, image):
+        from controlnet_aux.uniformer import UniformerSegmentor
+
         model = UniformerSegmentor.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
         del model

node_wrappers/zoe.py

@@ -1,5 +1,4 @@
 from ..utils import common_annotator_call, annotator_ckpts_path, HF_MODEL_NAME
-from controlnet_aux.zoe import ZoeDetector
 import comfy.model_management as model_management
 
 class Zoe_Depth_Map_Preprocessor:
@@ -14,6 +13,8 @@ def INPUT_TYPES(s):
     CATEGORY = "ControlNet Preprocessors/Normal and Depth Map"
 
     def execute(self, image, **kwargs):
+        from controlnet_aux.zoe import ZoeDetector
+
         model = ZoeDetector.from_pretrained(HF_MODEL_NAME, cache_dir=annotator_ckpts_path).to(model_management.get_torch_device())
         out = common_annotator_call(model, image)
         del model