Update app.py

app.py

@@ -12,7 +12,7 @@ blip_model = BlipForConditionalGeneration.from_pretrained(
     "Salesforce/blip-image-captioning-base"
 )
 
-# Hugging Face pipelines (temperature=1 for maximum diversity)
+# Hugging Face pipelines (all using Flan-T5-small for speed, temperature=1.0)
 category_generator = pipeline(
     "text2text-generation",
     model="google/flan-t5-small",
@@ -20,29 +20,25 @@ category_generator = pipeline(
     max_new_tokens=50,
     do_sample=True,
     temperature=1.0
-)  # Flan-T5-small for fast category generation
+)
+
 analysis_generator = pipeline(
     "text2text-generation",
     model="google/flan-t5-small",
     tokenizer="google/flan-t5-small",
-    max_new_tokens=500,  # increased from 200 to 500
-    do_sample=True,
-    temperature=1.0
-)  # Switched to Flan-T5-small for faster analysis  # reduced tokens for speed
+    max_new_tokens=500,
     do_sample=True,
     temperature=1.0
-)  # Switched to Flan-T5-small for faster analysis
+)
+
 suggestion_generator = pipeline(
     "text2text-generation",
     model="google/flan-t5-small",
     tokenizer="google/flan-t5-small",
-    max_new_tokens=500,  # increased from 200 to 500
+    max_new_tokens=500,
     do_sample=True,
     temperature=1.0
-)  # Using Flan-T5-small for quicker suggestions  # reduced tokens for speed
-    do_sample=True,
-    temperature=1.0
-)  # Using Flan-T5-small for quicker suggestions
+)
 
 # Example URLs for gallery
 def get_recommendations():
@@ -59,14 +55,13 @@ def get_recommendations():
         "https://i.imgur.com/Xj92Cjv.jpeg",
     ]
 
-# Generate raw BLIP caption
+# Generate BLIP caption from image
 def generate_caption(image):
     inputs = blip_processor(images=image, return_tensors="pt")
     outputs = blip_model.generate(**inputs)
     return blip_processor.decode(outputs[0], skip_special_tokens=True)
 
-# Derive concise category via Flan
-
+# Generate concise category via Flan
 def generate_category(caption):
     prompt = (
         f"Caption: {caption}\n"
@@ -76,7 +71,6 @@ def generate_category(caption):
     return raw.splitlines()[0]
 
 # Produce 5-sentence analysis via Flan
-
 def generate_analysis(caption):
     prompt = (
         f"Caption: {caption}\n"
@@ -87,7 +81,6 @@ def generate_analysis(caption):
     return " ".join(sentences[:5])
 
 # Suggest 5 bullet-point improvements via Flan
-
 def generate_suggestions(caption):
     prompt = (
         f"Caption: {caption}\n"
@@ -97,7 +90,7 @@ def generate_suggestions(caption):
     lines = [l for l in raw.splitlines() if l.strip().startswith('-')]
     return "\n".join(lines[:5]) if lines else "\n".join(raw.splitlines()[:5])
 
-# Full pipeline
+# Full pipeline combining all steps
 def process(image):
     caption = generate_caption(image)
     category = generate_category(caption)
@@ -106,11 +99,11 @@ def process(image):
     recs = get_recommendations()
     return category, analysis, suggestions, recs
 
-# UI Layout
+# UI Layout using Gradio
 with gr.Blocks(theme=gr.themes.Default(primary_hue="blue")) as demo:
     gr.Markdown("## 📢 Smart Ad Analyzer")
     gr.Markdown(
-        "Upload an image ad. BLIP captions it; Flan interprets that into Category, Analysis, and Suggestions."
+        "Upload an image ad to see: an Ad Category label, a five-sentence analysis, five bullet-point improvements, and example ads."
     )
 
     with gr.Row():
@@ -121,7 +114,6 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue="blue")) as demo:
             suggestion_out = gr.Textbox(label="Improvement Suggestions", lines=5, interactive=False)
             btn = gr.Button("Analyze Ad", size="sm", variant="primary")
 
-    # Full-width gallery below
     recommendation_gallery = gr.Gallery(label="Recommended Example Ads", show_label=True)
 
     btn.click(