Jacobmadwed commited on
Commit
28cd91d
·
verified ·
1 Parent(s): e80d315

Update handler.py

Browse files
Files changed (1) hide show
  1. handler.py +10 -12
handler.py CHANGED
@@ -9,7 +9,6 @@ import diffusers
9
  from diffusers.utils import load_image
10
  from diffusers.models import ControlNetModel
11
  from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
12
- from insightface.app import FaceAnalysis
13
  from style_template import styles
14
  from pipeline_stable_diffusion_xl_instantid_full import StableDiffusionXLInstantIDPipeline, draw_kps
15
  from controlnet_aux import OpenposeDetector
@@ -21,6 +20,7 @@ import base64
21
  import io
22
  import json
23
  from transformers import CLIPProcessor, CLIPModel
 
24
 
25
  # global variable
26
  device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -58,15 +58,13 @@ class EndpointHandler:
58
  controlnet_model = hf_hub_download(repo_id="InstantX/InstantID", filename="ControlNetModel/diffusion_pytorch_model.safetensors", local_dir=os.path.join(model_dir, "checkpoints"))
59
  face_adapter = hf_hub_download(repo_id="InstantX/InstantID", filename="ip-adapter.bin", local_dir=os.path.join(model_dir, "checkpoints"))
60
 
61
- dir_path = os.path.join(model_dir, "models", "antelopev2")
62
- if not os.path.exists(dir_path):
63
- print(f"Model path {dir_path} does not exist. Attempting to download.")
64
- self.app = FaceAnalysis(name='antelopev2', root=model_dir, providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
65
- else:
66
- print(f"Model path {dir_path} exists. Skipping download.")
67
- self.app = FaceAnalysis(name='antelopev2', root=model_dir, providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
68
 
69
- self.app.prepare(ctx_id=0, det_size=(640, 640))
70
  openpose = OpenposeDetector.from_pretrained("lllyasviel/ControlNet")
71
 
72
  # Path to InstantID models
@@ -235,8 +233,8 @@ class EndpointHandler:
235
  face_image_cv2 = convert_from_image_to_cv2(face_image)
236
  height, width, _ = face_image_cv2.shape
237
 
238
- # Extract face features
239
- face_info = self.app.get(face_image_cv2)
240
 
241
  face_info = sorted(
242
  face_info,
@@ -252,7 +250,7 @@ class EndpointHandler:
252
  img_controlnet = pose_image
253
  pose_image_cv2 = convert_from_image_to_cv2(pose_image)
254
 
255
- face_info = self.app.get(pose_image_cv2)
256
 
257
  face_info = face_info[-1]
258
  face_kps = draw_kps(pose_image, face_info["kps"])
 
9
  from diffusers.utils import load_image
10
  from diffusers.models import ControlNetModel
11
  from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
 
12
  from style_template import styles
13
  from pipeline_stable_diffusion_xl_instantid_full import StableDiffusionXLInstantIDPipeline, draw_kps
14
  from controlnet_aux import OpenposeDetector
 
20
  import io
21
  import json
22
  from transformers import CLIPProcessor, CLIPModel
23
+ import pickle
24
 
25
  # global variable
26
  device = "cuda" if torch.cuda.is_available() else "cpu"
 
58
  controlnet_model = hf_hub_download(repo_id="InstantX/InstantID", filename="ControlNetModel/diffusion_pytorch_model.safetensors", local_dir=os.path.join(model_dir, "checkpoints"))
59
  face_adapter = hf_hub_download(repo_id="InstantX/InstantID", filename="ip-adapter.bin", local_dir=os.path.join(model_dir, "checkpoints"))
60
 
61
+ # Load the new model
62
+ senet_model_path = os.path.join(model_dir, "models", "senet50.pkl")
63
+ if not os.path.exists(senet_model_path):
64
+ print(f"Model path {senet_model_path} does not exist. Please ensure the model is available.")
65
+ with open(senet_model_path, 'rb') as f:
66
+ self.senet_model = pickle.load(f)
 
67
 
 
68
  openpose = OpenposeDetector.from_pretrained("lllyasviel/ControlNet")
69
 
70
  # Path to InstantID models
 
233
  face_image_cv2 = convert_from_image_to_cv2(face_image)
234
  height, width, _ = face_image_cv2.shape
235
 
236
+ # Extract face features using the new model
237
+ face_info = self.senet_model.get(face_image_cv2)
238
 
239
  face_info = sorted(
240
  face_info,
 
250
  img_controlnet = pose_image
251
  pose_image_cv2 = convert_from_image_to_cv2(pose_image)
252
 
253
+ face_info = self.senet_model.get(pose_image_cv2)
254
 
255
  face_info = face_info[-1]
256
  face_kps = draw_kps(pose_image, face_info["kps"])