Update app.py

app.py

@@ -1,54 +1,52 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
-import PIL.Image
-import io
-import base64
+import torch
+from diffusers import StableDiffusionControlNetPipeline, ControlNetModel
+from PIL import Image
+import numpy as np
 
-# Initialize the Hugging Face client
-HF_TOKEN = "your_token_here"  # Replace with your token
+def load_model():
+    controlnet = ControlNetModel.from_pretrained("Kwai-Kolors/Kolors-Virtual-Try-On")
+    pipe = StableDiffusionControlNetPipeline.from_pretrained(
+        "Kwai-Kolors/Kolors-Virtual-Try-On",
+        controlnet=controlnet,
+        torch_dtype=torch.float16
+    )
+    if torch.cuda.is_available():
+        pipe = pipe.to("cuda")
+    return pipe
 
-client = InferenceClient(
-    model="Kwai-Kolors/Kolors-Virtual-Try-On",
-    token=HF_TOKEN
-)
+# Model'i global olarak yükle
+try:
+    model = load_model()
+    print("Model başarıyla yüklendi!")
+except Exception as e:
+    print(f"Model yüklenirken hata: {str(e)}")
 
 def virtual_try_on(person_image, garment_image):
     """
-    Process the virtual try-on request
-    Args:
-        person_image: PIL Image of the person
-        garment_image: PIL Image of the garment
-    Returns:
-        PIL Image of the result
+    Virtual try-on process
     """
     try:
-        # Convert images to bytes
-        person_bytes = io.BytesIO()
-        garment_bytes = io.BytesIO()
-        person_image.save(person_bytes, format='PNG')
-        garment_image.save(garment_bytes, format='PNG')
-        
-        # Convert bytes to base64
-        person_b64 = base64.b64encode(person_bytes.getvalue()).decode('utf-8')
-        garment_b64 = base64.b64encode(garment_bytes.getvalue()).decode('utf-8')
-        
-        # Make API request
-        response = client.post(
-            json={
-                "inputs": [
-                    {"image": person_b64},
-                    {"image": garment_b64}
-                ]
-            }
+        # Resimleri uygun formata dönüştür
+        if person_image is None or garment_image is None:
+            return None, "Error: Both images are required"
+            
+        # Model inference
+        output = model(
+            person_image,
+            garment_image,
+            num_inference_steps=30,
+            guidance_scale=7.5
         )
         
-        # Convert response to image
-        result_image = PIL.Image.open(io.BytesIO(response))
+        # Sonuç resmini al
+        result_image = output.images[0]
         return result_image, "Success"
+        
     except Exception as e:
         return None, f"Error: {str(e)}"
 
-# Create Gradio interface
+# Gradio arayüzü
 demo = gr.Interface(
     fn=virtual_try_on,
     inputs=[
@@ -59,7 +57,7 @@ demo = gr.Interface(
         gr.Image(type="pil", label="Result"),
         gr.Text(label="Status")
     ],
-    title="Virtual Try-On API",
+    title="Virtual Try-On",
     description="Upload a person image and a garment image to see how the garment would look on the person."
 )