Update app.py

app.py

@@ -1,61 +1,66 @@
-import gradio as gr
+pip install gradio transformers torch Pillow
+import gradio as gr 
 from transformers import MarianMTModel, MarianTokenizer, BlipProcessor, BlipForConditionalGeneration
 from PIL import Image
 import torch
 
-# Load the Tamil-to-English translation model
-model_name = "Helsinki-NLP/opus-mt-ta-en"
-tokenizer = MarianTokenizer.from_pretrained(model_name)
-translation_model = MarianMTModel.from_pretrained(model_name)
+# 1. Load Tamil to English translation model
+translation_model_name = "Helsinki-NLP/opus-mt-ta-en"
+tokenizer = MarianTokenizer.from_pretrained(translation_model_name)
+translation_model = MarianMTModel.from_pretrained(translation_model_name)
 
-# Load the BLIP model for image captioning
+# 2. Load BLIP model for image captioning
 caption_processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
 caption_model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
 
+# 3. Translation function
 def translate_tamil_to_english(tamil_text):
-    inputs = tokenizer(tamil_text, return_tensors="pt", padding=True)
+    if not tamil_text.strip():
+        return "No input given"
+    inputs = tokenizer(tamil_text, return_tensors="pt", padding=True, truncation=True)
     translated = translation_model.generate(**inputs)
     english_text = tokenizer.decode(translated[0], skip_special_tokens=True)
     return english_text
 
-# Generate image using text (stub – replace with actual model if needed)
+# 4. Generate dummy image from text
 def generate_image_from_text(text_prompt):
-    # Instead of using Stable Diffusion, just show a sample image
-    img = Image.new('RGB', (512, 512), color='lightblue')
+    # Use a plain color image with PIL
+    img = Image.new('RGB', (512, 512), color=(200, 230, 255))
     return img
 
+# 5. Describe the image
 def describe_image(image):
     inputs = caption_processor(images=image, return_tensors="pt")
     out = caption_model.generate(**inputs)
     caption = caption_processor.decode(out[0], skip_special_tokens=True)
     return caption
 
+# 6. Combined pipeline
 def full_pipeline(tamil_text):
     english_text = translate_tamil_to_english(tamil_text)
     generated_image = generate_image_from_text(english_text)
-    description = describe_image(generated_image)
-    return english_text, generated_image, description
+    caption = describe_image(generated_image)
+    return english_text, generated_image, caption
 
-# Gradio interface
+# 7. Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("## Tamil to English → Image → Description")
+    gr.Markdown("## 🌍 Tamil ➝ English ➝ Image ➝ Description App")
 
     with gr.Row():
-        tamil_input = gr.Textbox(label="Enter Tamil Text", lines=2, placeholder="உதாரணம்: ஒரு பூந்தோட்டத்தில் செருப்புகள் இருக்கின்றன")
-    
-    with gr.Row():
-        translate_btn = gr.Button("Translate and Generate")
+        tamil_input = gr.Textbox(label="Enter Tamil Text", lines=2, placeholder="உதாரணம்: ஒரு பூங்காவில் ஒரு பசுமை மரம் உள்ளது")
+
+    translate_btn = gr.Button("Translate, Generate Image, and Describe")
 
     with gr.Row():
-        english_output = gr.Textbox(label="Translated English Text")
-        description_output = gr.Textbox(label="Image Description")
+        english_output = gr.Textbox(label="Translated English")
+        caption_output = gr.Textbox(label="Image Description")
 
     image_output = gr.Image(label="Generated Image")
 
     translate_btn.click(
         fn=full_pipeline,
         inputs=tamil_input,
-        outputs=[english_output, image_output, description_output]
+        outputs=[english_output, image_output, caption_output]
     )
 
 demo.launch()