Update app.py

app.py

@@ -1,54 +1,69 @@
 import gradio as gr
 import numpy as np
 import random
-import spaces #[uncomment to use ZeroGPU]
+import spaces
 from diffusers import DiffusionPipeline, DPMSolverSDEScheduler
 import torch
 from huggingface_hub import hf_hub_download
 from ultralytics import YOLO
-import cv2
 from PIL import Image
+import cv2
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
-model_repo_id = "John6666/wai-ani-nsfw-ponyxl-v8-sdxl"  # Your diffusion model
+model_repo_id = "John6666/wai-ani-nsfw-ponyxl-v8-sdxl"
+adetailer_model_id = "Bingsu/adetailer"  # Your ADetailer model
+
+# Load the YOLO model for face detection
+yolo_model_path = hf_hub_download(adetailer_model_id, "face_yolov8n.pt")
+yolo_model = YOLO(yolo_model_path)
+
+if torch.cuda.is_available():
+    torch_dtype = torch.float16
+else:
+    torch_dtype = torch.float32
 
-# Load your main diffusion pipeline
-pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch.float16)
+pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
 pipe.scheduler = DPMSolverSDEScheduler.from_config(pipe.scheduler.config, algorithm_type="dpmsolver++", solver_order=2, use_karras_sigmas=True)
 pipe = pipe.to(device)
 
 MAX_SEED = np.iinfo(np.int32).max
 MAX_IMAGE_SIZE = 1024
 
-# Download the ADetailer YOLOv8 face detection model
-yolo_model_path = hf_hub_download(repo_id="Bingsu/adetailer", filename="face_yolov8n.pt")
-yolo_model = YOLO(yolo_model_path)
-
-def fix_eyes_with_adetailer(image):
-    # Convert PIL image to OpenCV format for YOLO
+def correct_anime_face(image):
+    # Convert to OpenCV format
     img = np.array(image)
     img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
     
-    # Run the YOLO model on the image
+    # Detect faces
     results = yolo_model(img)
     
-    # Visualize and process the output
-    pred = results[0].plot()  # Draw bounding boxes and other detections
-    pred = cv2.cvtColor(pred, cv2.COLOR_BGR2RGB)
+    for detection in results[0].boxes:
+        x1, y1, x2, y2 = map(int, detection.xyxy[0].tolist())
+        
+        # Crop the face region
+        face = img[y1:y2, x1:x2]
+        face_pil = Image.fromarray(cv2.cvtColor(face, cv2.COLOR_BGR2RGB))
+        
+        # Prompt for the correction model
+        prompt = "Enhance this anime character's face, fix eyes and make features more vivid."
+        
+        # Process the face with the anime correction model
+        corrected_face = pipe(prompt=prompt, image=face_pil).images[0]  # Replace with your correction model
+        
+        # Place the corrected face back into the original image
+        img[y1:y2, x1:x2] = np.array(corrected_face)
     
-    # Convert the processed image back to PIL format
-    corrected_image = Image.fromarray(pred)
-    return corrected_image
+    # Convert back to PIL
+    final_image = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+    return final_image
 
-@spaces.GPU #[uncomment to use ZeroGPU]
+@spaces.GPU
 def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
-
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
         
     generator = torch.Generator().manual_seed(seed)
     
-    # Generate the initial image with the diffusion model
     image = pipe(
         prompt=prompt, 
         negative_prompt=negative_prompt,
@@ -58,9 +73,9 @@ def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance
         height=height,
         generator=generator
     ).images[0] 
-
-    # Apply ADetailer to fix the eyes after generating the image
-    corrected_image = fix_eyes_with_adetailer(image)
+    
+    # Correct anime face in the generated image
+    corrected_image = correct_anime_face(image)
     
     return corrected_image, seed
 
@@ -70,94 +85,41 @@ examples = [
     "A delicious ceviche cheesecake slice",
 ]
 
-css="""#col-container {margin: 0 auto; max-width: 640px;}"""
+css = """
+#col-container {
+    margin: 0 auto;
+    max-width: 640px;
+}
+"""
 
 with gr.Blocks(css=css) as demo:
-    
     with gr.Column(elem_id="col-container"):
-        gr.Markdown(f"""
-        # Text-to-Image Gradio Template
-        """)
+        gr.Markdown("# Text-to-Image Gradio Template")
         
         with gr.Row():
-            
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-            
+            prompt = gr.Text(label="Prompt", show_label=False, max_lines=1, placeholder="Enter your prompt", container=False)
             run_button = gr.Button("Run", scale=0)
         
         result = gr.Image(label="Result", show_label=False)
 
         with gr.Accordion("Advanced Settings", open=False):
-            
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
-            )
-            
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-            
+            negative_prompt = gr.Text(label="Negative prompt", max_lines=1, placeholder="Enter a negative prompt", visible=False)
+            seed = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0)
             randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
             
             with gr.Row():
+                width = gr.Slider(label="Width", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
+                height = gr.Slider(label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
                 
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024, #Replace with defaults that work for your model
-                )
-                
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024, #Replace with defaults that work for your model
-                )
-            
             with gr.Row():
-                
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=10.0,
-                    step=0.1,
-                    value=0.0, #Replace with defaults that work for your model
-                )
-                
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=50,
-                    step=1,
-                    value=2, #Replace with defaults that work for your model
-                )
+                guidance_scale = gr.Slider(label="Guidance scale", minimum=0.0, maximum=10.0, step=0.1, value=0.0)
+                num_inference_steps = gr.Slider(label="Number of inference steps", minimum=1, maximum=50, step=1, value=2)
         
-        gr.Examples(
-            examples=examples,
-            inputs=[prompt]
-        )
+        gr.Examples(examples=examples, inputs=[prompt])
     
-    gr.on(
-        triggers=[run_button.click, prompt.submit],
-        fn=infer,
-        inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
-        outputs=[result, seed]
-    )
+    gr.on(triggers=[run_button.click, prompt.submit],
+          fn=infer,
+          inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
+          outputs=[result, seed])
 
 demo.queue().launch()