Update app.py

app.py

@@ -92,14 +92,7 @@ def get_EOS_pos_in_prompt(prompt):
 
 
 import torch.nn.functional as F
-"""
-def gradient_loss(images):
-    # Compute gradient magnitude using Sobel filters.
-    gradient_x = F.conv2d(images, torch.Tensor([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]]).view(1, 1, 3, 3).to(images.device))
-    gradient_y = F.conv2d(images, torch.Tensor([[-1, -2, -1], [0, 0, 0], [1, 2, 1]]).view(1, 1, 3, 3).to(images.device))
-    gradient_magnitude = torch.sqrt(gradient_x**2 + gradient_y**2)
-    return gradient_magnitude.mean()
-"""
+
 
 from torchvision.transforms import ToTensor
 def pil_to_latent(input_im):
@@ -317,13 +310,13 @@ def image_generator(prompt="cat", loss_function=None):
         images_without_loss.append(generated_img)
 
         if loss_function:
-            if loss_function == "exposure_loss":
+            if loss_function == "Exposure":
                 generated_img = generate_image_custom_style(prompt, style_num=i, random_seed=seed_values[i], custom_loss_fn=exposure_loss)
-            elif loss_function == "color_diversity_loss":
+            elif loss_function == "Color Diversity":
                 generated_img = generate_image_custom_style(prompt, style_num=i, random_seed=seed_values[i], custom_loss_fn=color_diversity_loss)
-            elif loss_function == "sharpness_loss":
+            elif loss_function == "Sharpness":
                 generated_img = generate_image_custom_style(prompt, style_num=i, random_seed=seed_values[i], custom_loss_fn=sharpness_loss)
-            elif loss_function == "brilliance_loss":
+            elif loss_function == "Brilliance":
                 generated_img = generate_image_custom_style(prompt, style_num=i, random_seed=seed_values[i], custom_loss_fn=brilliance_loss)
             images_with_loss.append(generated_img)
 
@@ -341,12 +334,39 @@ def image_generator(prompt="cat", loss_function=None):
 def image_generator_wrapper(prompt="dog", selected_loss="None"):
     return image_generator(prompt, selected_loss)
 
-description = 'Stable Diffusion is a generative artificial intelligence (generative AI) model that produces unique photorealistic images from text and image prompts.'
-title = 'Image Generation using Stable Diffusion'
+icon_html = '<i class="fas fa-chart-bar"></i>'
+title_with_icon = f"""
+<div style="background-color: #f5f1f2; padding: 10px; display: flex; align-items: center;">
+    {icon_html} <span style="margin-left: 10px;">Image Generation using Stable Diffusion</span>
+</div>
+"""
+description_with_icon = f"""
+<div style="background-color: #f1f1f5; padding: 10px; display: flex; align-items: center;">
+    {icon_html}
+    <span style="margin-left: 10px;">
+        <p><strong>Embedding New Styles Into Stable Diffusion</strong></p>
+        <p><strong>Following are the concepts trained on</strong></p>
+        <ul>
+            <li>animal-toy</li>
+            <li>fft</li>
+            <li>midjourney</li>
+            <li>oil style</li>
+            <li>space style</li>
+        </ul>
+        <p>Following are some Losses tried</p>
+        <ul>
+            <li>exposure : It helps control the overall exposure of generated images. It ensures that the contrast of the generated images align with the desired aesthetic, preventing overexposure or underexposure</li>
+            <li>Brilliance:  Brilliance loss is a loss function that emphasizes the brilliance or luminance of specific image components, such as highlights. It can be used to highlight or enhance certain aspects of the generated artwork, adding a touch of brilliance or radiance to the final image.</li>
+            <li>color diversity: Color diversity loss encourages the model to produce images with a wider range of colors and hues. It helps create visually diverse and vibrant artworks by minimizing color repetition and promoting a rich color palette in the generated images</li>
+            <li>sharpness: Sharpness loss is used to enhance the level of detail and clarity in generated images. It encourages the model to produce crisp and well-defined visual elements, leading to sharper and more realistic results.</li>
+        </ul>
+    </span>
+</div>
+"""
 
 demo = gr.Interface(image_generator_wrapper,
-                    inputs=[gr.Textbox(label="Enter prompt for generation", type="text", value="A ballerina cat dancing in space"),
-                            gr.Radio(["None", "exposure_loss", "color_diversity_loss", "sharpness_loss", "brilliance_loss"], value="None", label="Select Loss")],
+                    inputs=[gr.Textbox(label="Enter prompt for generating Image", type="text", value="A ballerina cat dancing in space"),
+                            gr.Radio(["None", "Exposure", "Color Diversity", "Sharpness", "Brilliance"], value="None", label="Select Loss")],
                     outputs=gr.Plot(label="Generated Images"),
                     title=title,
                     description=description)