create readme

src/app_cached_videos.py

@@ -44,7 +44,7 @@ def overlay_mask(img, mask):
     alpha_channel[mask_resized < 5] = 0
 
     # Set the opacity level for non-black pixels
-    opacity = 0.3  # Adjust this value as needed (0.0 to 1.0)
+    opacity = 0.35  # Adjust this value as needed (0.0 to 1.0)
     alpha_channel[mask_resized != 0] = int(255 * opacity)
 
     # Create a 4-channel image (BGR + Alpha)
@@ -106,9 +106,6 @@ data_path = Path(
 available_prompts = ['a dog and a cat sitting','A fish swimming in the water', 'A spider descending from its web', 'An astronaut riding a horse']
 
 with gr.Blocks(css=css) as demo:
-    # gr.Markdown("""
-    #             # Peekaboo Demo
-    #             """)
     with gr.Row():
         video_1 = gr.Image(label="Image", )
         # video_1 = gr.Image(label="Image", width=256, height=256)
@@ -120,6 +117,14 @@ with gr.Blocks(css=css) as demo:
         return gr.Radio(text.strip().split(' '), value=None, label='Choose a word to visualize its attention mask.')
 
     text = 'a dog and a cat sitting'
+    gr.Markdown("""
+                ## Visualizing Attention Masks
+                * Select a prompt from the drop down
+                * Click on "Get words" to get the words in the prompt
+                * Select a radio button from the words to visualize the attention mask
+                * Play around with the index of diffusion steps, layers to visualize different masks
+                """)
+
     with gr.Group("Video Selection"):
         txt_1 = gr.Dropdown(choices=available_prompts, label="Video Prompt", value=available_prompts[0])
         submit_btn = gr.Button('Get words')