Upload web-ui.py

web-ui.py

@@ -14,6 +14,7 @@ from pyngrok import ngrok
 import threading
 import time
 from ip_adapter.ip_adapter_faceid import IPAdapterFaceIDXL
+import hashlib
 
 # Argument parser for command line options
 parser = argparse.ArgumentParser()
@@ -51,6 +52,12 @@ static_model_names = [
 model_cache = {}
 max_cache_size = args.cache_limit
 
+embeddings_cache = {}
+
+def get_image_hash(image):
+    image_bytes = image.tobytes()
+    return hashlib.sha256(image_bytes).hexdigest()
+
 def convert_model(checkpoint_path, output_path, isSDXL):
     try:
         if isSDXL:
@@ -126,19 +133,27 @@ def generate_image(input_image, positive_prompt, negative_prompt, width, height,
     # Load and prepare the model
     ip_model = load_model(model_name, isSDXL)
 
-    # Convert input image to the format expected by the model
+    # Convert input image to the format expected by the model and calculate its hash
     input_image = input_image.convert("RGB")
-    input_image = cv2.cvtColor(np.array(input_image), cv2.COLOR_RGB2BGR)
-    app = FaceAnalysis(
-        name="buffalo_l", providers=["CUDAExecutionProvider", "CPUExecutionProvider"]
-    )
-    app.prepare(ctx_id=0, det_size=(640, 640))
-    faces = app.get(input_image)
-    if not faces:
-        raise ValueError("No faces found in the image.")
+    input_image_cv2 = cv2.cvtColor(np.array(input_image), cv2.COLOR_RGB2BGR)
+    image_hash = get_image_hash(input_image)
 
-    faceid_embeds = torch.from_numpy(faces[0].normed_embedding).unsqueeze(0)
-    face_image = face_align.norm_crop(input_image, landmark=faces[0].kps, image_size=224)
+    # Check if embeddings are cached
+    if image_hash in embeddings_cache:
+        faceid_embeds, face_image = embeddings_cache[image_hash]
+    else:
+        app = FaceAnalysis(
+            name="buffalo_l", providers=["CUDAExecutionProvider", "CPUExecutionProvider"]
+        )
+        app.prepare(ctx_id=0, det_size=(640, 640))
+        faces = app.get(input_image_cv2)
+        if not faces:
+            raise ValueError("No faces found in the image.")
+
+        faceid_embeds = torch.from_numpy(faces[0].normed_embedding).unsqueeze(0)
+        face_image = face_align.norm_crop(input_image_cv2, landmark=faces[0].kps, image_size=224)
+        # Cache the embeddings
+        embeddings_cache[image_hash] = (faceid_embeds, face_image)
 
     for image_index in range(num_images):
         if randomize_seed or image_index > 0: