Update app.py

app.py

@@ -10,9 +10,10 @@ from transformers import (
     AutoModelForSeq2SeqLM,
 )
 
+# Auto-detect CPU/GPU
 DEVICE = 0 if torch.cuda.is_available() else -1
 
-# BLIP
+# BLIP captioner
 processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
 blip_model = AutoModelForVision2Seq.from_pretrained("Salesforce/blip-image-captioning-large")
 caption_pipe = pipeline(
@@ -23,6 +24,7 @@ caption_pipe = pipeline(
     device=DEVICE,
 )
 
+# FLAN-T5 for text-to-text
 FLAN_MODEL = "google/flan-t5-large"
 flan_tokenizer = AutoTokenizer.from_pretrained(FLAN_MODEL)
 flan_model = AutoModelForSeq2SeqLM.from_pretrained(FLAN_MODEL)
@@ -36,6 +38,7 @@ category_pipe = pipeline(
     do_sample=True,
     temperature=1.0,
 )
+
 analysis_pipe = pipeline(
     "text2text-generation",
     model=flan_model,
@@ -45,6 +48,7 @@ analysis_pipe = pipeline(
     do_sample=True,
     temperature=1.0,
 )
+
 suggestion_pipe = pipeline(
     "text2text-generation",
     model=flan_model,
@@ -54,6 +58,7 @@ suggestion_pipe = pipeline(
     do_sample=True,
     temperature=1.0,
 )
+
 expansion_pipe = pipeline(
     "text2text-generation",
     model=flan_model,
@@ -80,18 +85,26 @@ def get_recommendations():
 def process(image: Image):
     if image is None:
         return "", "", "", "", get_recommendations()
+
+    # BLIP caption
     caption_res = caption_pipe(image, max_new_tokens=64)
     raw_caption = caption_res[0]["generated_text"].strip()
+
+    # Expand if too short
     if len(raw_caption.split()) < 3:
         exp = expansion_pipe(f"Expand into a detailed description: {raw_caption}")
         desc = exp[0]["generated_text"].strip()
     else:
         desc = raw_caption
+
+    # Category
     cat_prompt = (
         f"Description: {desc}\n\n"
         "Provide a concise category label for this ad (e.g. 'Food', 'Fitness'):"
     )
     cat_out = category_pipe(cat_prompt)[0]["generated_text"].splitlines()[0].strip()
+
+    # Five-sentence analysis
     ana_prompt = (
         f"Description: {desc}\n\n"
         "Write exactly five sentences explaining what this ad communicates and its emotional impact."
@@ -99,67 +112,68 @@ def process(image: Image):
     ana_raw = analysis_pipe(ana_prompt)[0]["generated_text"].strip()
     sentences = re.split(r'(?<=[.!?])\s+', ana_raw)
     analysis = " ".join(sentences[:5])
+
+    # Five bullet-point suggestions - unique, high-quality
     sug_prompt = (
         f"Description: {desc}\n\n"
-        "Suggest five **different** and actionable improvements for this ad. "
-        "Each must start with '- ' and be a single sentence. "
-        "Avoid repeating any idea or wording."
+        "Provide five distinct improvement suggestions for this ad. Each must start with '- ', be one sentence, and not repeat the same idea."
     )
     sug_raw = suggestion_pipe(sug_prompt)[0]["generated_text"].strip()
-    seen = set()
     bullets = []
+    seen = set()
     for l in sug_raw.splitlines():
-        if l.startswith("-"):
-            key = l[2:].strip().lower()
-            if key and key not in seen:
+        line = l.strip()
+        if line.startswith("-"):
+            key = line[2:].lower()
+            if key not in seen and len(key) > 5:
+                bullets.append(line)
                 seen.add(key)
-                bullets.append(l.strip())
         if len(bullets) == 5:
             break
+    # Fallbacks
     fallback = [
         "- Add a bold and visible call-to-action button.",
         "- Use brighter colors or higher contrast for more visual impact.",
         "- Refine the text for greater clarity and conciseness.",
         "- Adjust the image layout for better balance and focus.",
-        "- Highlight product benefits more clearly in the headline.",
+        "- Highlight product benefits more clearly in the headline."
     ]
     for fb in fallback:
         if len(bullets) == 5:
             break
-        fb_key = fb[2:].strip().lower()
+        fb_key = fb[2:].lower()
         if fb_key not in seen:
             bullets.append(fb)
             seen.add(fb_key)
     suggestions = "\n".join(bullets[:5])
-    return "", cat_out, analysis, suggestions, get_recommendations()
+
+    return "", cat_out, analysis, suggestions, get_recommendations()  # Hides BLIP Caption by returning blank
 
 def main():
     with gr.Blocks(title="Smart Ad Analyzer") as demo:
-        gr.Markdown("#  Smart Ad Analyzer")
         gr.Markdown(
-            """
-            Upload an ad image and get AI-powered creative feedback instantly:
-            -  **Ad Category** (concise and relevant)
-            -  **Five-sentence Analysis** (ad message, design, impact)
-            -  **Five unique Improvement Suggestions**
-            -  **Example Ads** for inspiration
-
-            Get quick, actionable advice for better ads—no creative block, no guesswork.
-            """
+            "##  Smart Ad Analyzer\n"
+            "Upload any advertisement image below and get an instant breakdown:\n\n"
+            "- **Ad Category:** Instantly identifies what sector your ad fits into\n"
+            "- **Ad Analysis:** Five concise sentences explaining what your ad communicates and the emotional response it targets\n"
+            "- **Improvement Suggestions:** Five practical, unique tips to make your ad more effective\n"
+            "- **Example Ads Gallery:** See proven ad designs for inspiration\n\n"
+            "_Ideal for marketers, business owners, students, and creative teams who want to boost ad impact using AI. No technical skill required._"
         )
         with gr.Row():
             inp = gr.Image(type='pil', label='Upload Ad Image')
             with gr.Column():
-                # BLIP caption hidden!
-                cat_out = gr.Textbox(label='📂 Ad Category', interactive=False)
-                ana_out = gr.Textbox(label='📊 Ad Analysis', lines=5, interactive=False)
-                sug_out = gr.Textbox(label='🚀 Improvement Suggestions', lines=8, interactive=False)
+                # To hide BLIP caption, don't add the textbox output
+                # cap_out = gr.Textbox(label=' BLIP Caption', interactive=False, visible=False)
+                cat_out = gr.Textbox(label=' Ad Category', interactive=False)
+                ana_out = gr.Textbox(label=' Ad Analysis', lines=5, interactive=False)
+                sug_out = gr.Textbox(label=' Improvement Suggestions', lines=5, interactive=False)
                 btn = gr.Button('Analyze Ad', variant='primary')
         gallery = gr.Gallery(label='Example Ads')
         btn.click(
             fn=process,
             inputs=[inp],
-            outputs=[cat_out, ana_out, sug_out, gallery],
+            outputs=["textbox", cat_out, ana_out, sug_out, gallery],  # BLIP output is blank, textbox hidden
         )
         gr.Markdown('Made by Simon Thalmay')
     return demo