Update app.py

app.py

@@ -1,48 +1,41 @@
-import os
-import cv2
 import gradio as gr
 import numpy as np
 import random
-import time
+import torch
+from diffusers import StableDiffusionPipeline
+
+# Load the Stable Diffusion model for text-based garment generation
+model_id = "runwayml/stable-diffusion-v1-5"
+pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
+pipe = pipe.to("cuda")  # Use GPU for faster inference
+
+MAX_SEED = 999999
+
+def generate_garment(person_img, cloth_description, seed, randomize_seed):
+    if person_img is None or cloth_description is None or cloth_description.strip() == "":
+        return None, None, "Invalid input"
 
-def tryon(person_img, garment_prompt, seed, randomize_seed):
-    post_start_time = time.time()
-    
-    if person_img is None or garment_prompt.strip() == "":
-        return None, None, "Empty image or prompt"
-    
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
-        
-    # Create a copy of the person image to overlay text
-    result_img = person_img.copy()
-    
-    # Convert the image to OpenCV format (if needed)
-    if len(result_img.shape) == 2:  # Convert grayscale to RGB
-        result_img = cv2.cvtColor(result_img, cv2.COLOR_GRAY2RGB)
-    
-    # Set text position and properties
-    text_position = (10, 30)
-    font = cv2.FONT_HERSHEY_SIMPLEX
-    font_scale = 1
-    font_color = (0, 255, 0)  # Green color for the text
-    thickness = 2
 
-    # Overlay the garment description text on the image
-    cv2.putText(result_img, f'Garment: {garment_prompt}', text_position, font, font_scale, font_color, thickness, cv2.LINE_AA)
+    # Generate garment image from the text description
+    torch.manual_seed(seed)
+    garment_img = pipe(cloth_description).images[0]
 
-    post_end_time = time.time()
-    print(f"post time used: {post_end_time - post_start_time}")
+    # Combine the generated garment with the person's image
+    result_img = combine_images(person_img, garment_img)
 
-    # Return the resulting image, used seed, and success message
     return result_img, seed, "Success"
 
-MAX_SEED = 999999
+def combine_images(person_img, garment_img):
+    person_img = np.array(person_img)
+    garment_img = np.array(garment_img.resize((person_img.shape[1], person_img.shape[0])))
 
-example_path = os.path.join(os.path.dirname(__file__), 'assets')
+    # Simple overlay of garment on the person image
+    # Further improvement may require segmentation/masking
+    result_img = np.where(garment_img[:, :, 3:] > 0, garment_img[:, :, :3], person_img)
 
-human_list = os.listdir(os.path.join(example_path, "human"))
-human_list_path = [os.path.join(example_path, "human", human) for human in human_list]
+    return result_img
 
 css = """
 #col-left {
@@ -61,90 +54,32 @@ css = """
     margin: 0 auto;
     max-width: 1100px;
 }
-#button {
-    color: blue;
-}
 """
 
-def load_description(fp):
-    with open(fp, 'r', encoding='utf-8') as f:
-        content = f.read()
-    return content
-
-
 with gr.Blocks(css=css) as Tryon:
-    gr.HTML(load_description("assets/title.md"))
-    with gr.Row():
-        with gr.Column(elem_id="col-left"):
-            gr.HTML("""
-            <div style="display: flex; justify-content: center; align-items: center; text-align: center; font-size: 20px;">
-                <div>
-                Step 1.  Upload a person image ⬇️
-                </div>
-            </div>
-            """)
-        with gr.Column(elem_id="col-mid"):
-            gr.HTML("""
-            <div style="display: flex; justify-content: center; align-items: center; text-align: center; font-size: 20px;">
-                <div>
-                Step 2. Enter a text prompt for the garment ⬇️
-                </div>
-            </div>
-            """)
-        with gr.Column(elem_id="col-right"):
-            gr.HTML("""
-            <div style="display: flex; justify-content: center; align-items: center; text-align: center; font-size: 20px;">
-                <div>
-                Step 3. Press “Run” to get try-on results
-                </div>
-            </div>
-            """)
+    gr.HTML("<h1>Virtual Try-On with Text-based Garment Generation</h1>")
+    
     with gr.Row():
         with gr.Column(elem_id="col-left"):
-            imgs = gr.Image(label="Person image", sources='upload', type="numpy")
-            example = gr.Examples(
-                inputs=imgs,
-                examples_per_page=12,
-                examples=human_list_path
-            )
+            gr.HTML("<h3>Step 1: Upload a person image ⬇️</h3>")
+            person_img = gr.Image(label="Person Image", source='upload', type="numpy")
+        
         with gr.Column(elem_id="col-mid"):
-            garment_prompt = gr.Textbox(label="Garment text prompt", placeholder="Describe the garment...")
+            gr.HTML("<h3>Step 2: Describe the garment ⬇️</h3>")
+            cloth_description = gr.Textbox(label="Garment Description", placeholder="e.g., red dress with floral pattern")
+        
         with gr.Column(elem_id="col-right"):
-            image_out = gr.Image(label="Result", show_share_button=False)
-            with gr.Row():
-                seed = gr.Slider(
-                    label="Seed",
-                    minimum=0,
-                    maximum=MAX_SEED,
-                    step=1,
-                    value=0,
-                )
-                randomize_seed = gr.Checkbox(label="Random seed", value=True)
-            with gr.Row():
-                seed_used = gr.Number(label="Seed used")
-                result_info = gr.Text(label="Response")
-            test_button = gr.Button(value="Run", elem_id="button")
-
-    test_button.click(fn=tryon, inputs=[imgs, garment_prompt, seed, randomize_seed], outputs=[image_out, seed_used, result_info], concurrency_limit=40)
-
-    with gr.Column(elem_id="col-showcase"):
-        gr.HTML("""
-        <div style="display: flex; justify-content: center; align-items: center; text-align: center; font-size: 20px;">
-            <div> </div>
-            <br>
-            <div>
-            Virtual try-on examples in pairs of person and garment images
-            </div>
-        </div>
-        """)
-        show_case = gr.Examples(
-            examples=[
-                ["assets/examples/model2.png", "assets/examples/garment2.png", "assets/examples/result2.png"],
-                ["assets/examples/model3.png", "assets/examples/garment3.png", "assets/examples/result3.png"],
-                ["assets/examples/model1.png", "assets/examples/garment1.png", "assets/examples/result1.png"],
-            ],
-            inputs=[imgs, garment_prompt, image_out],
-            label=None
-        )
+            gr.HTML("<h3>Step 3: Generate Try-On Image ⬇️</h3>")
+            result_img = gr.Image(label="Result", show_share_button=False)
+            seed = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0)
+            randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
+            seed_used = gr.Number(label="Seed Used", interactive=False)
+            result_info = gr.Text(label="Status", interactive=False)
+            
+            generate_button = gr.Button(value="Run")
+        
+    generate_button.click(fn=generate_garment, 
+                          inputs=[person_img, cloth_description, seed, randomize_seed], 
+                          outputs=[result_img, seed_used, result_info])
 
 Tryon.launch()