Update app.py

app.py

@@ -1,33 +1,37 @@
 import gradio as gr
-from transformers import BlipProcessor, BlipForConditionalGeneration
+import tensorflow as tf
+from tensorflow.keras.applications import ResNet50
+from tensorflow.keras.applications.resnet50 import preprocess_input, decode_predictions
 from PIL import Image
-import torch
+import numpy as np
 
-# Load the Microsoft Phi-3.5-mini-instruct model
-model_name = "microsoft/phi-3.5-mini-instruct"
-processor = BlipProcessor.from_pretrained(model_name)
-model = BlipForConditionalGeneration.from_pretrained(model_name)
+# Load pre-trained ResNet50 model + higher level layers
+model = ResNet50(weights='imagenet')
 
-# Define the image classification function
-def classify_image(image):
+def classify_image(img):
+    # Resize the image to 224x224 pixels (required input size for ResNet50)
+    img = img.resize((224, 224))
+    # Convert the image to array
+    img_array = np.array(img)
+    # Expand dimensions to match the shape required by the model
+    img_array = np.expand_dims(img_array, axis=0)
     # Preprocess the image
-    inputs = processor(images=image, return_tensors="pt")
-    pixel_values = inputs["pixel_values"]
-
-    # Generate the classification
-    with torch.no_grad():
-        generated_ids = model.generate(pixel_values=pixel_values)
-    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+    img_array = preprocess_input(img_array)
+    # Predict the classification
+    predictions = model.predict(img_array)
+    # Decode predictions into readable labels
+    decoded_predictions = decode_predictions(predictions, top=3)[0]
     
-    return generated_text
+    # Format the output
+    return {label: float(probability) for (_, label, probability) in decoded_predictions}
 
-# Create a Gradio interface
+# Gradio interface
 iface = gr.Interface(
-    fn=classify_image,
-    inputs=gr.Image(type="pil"),
-    outputs=gr.Textbox(label="Image Classification"),
-    title="Image Context Classification",
-    description="Upload an image to classify its context using Microsoft's Phi-3.5-mini-instruct model."
+    fn=classify_image,  # Function to call for predictions
+    inputs=gr.Image(type="pil"),  # Input is an image
+    outputs=gr.Label(num_top_classes=3),  # Output is a label with top 3 predictions
+    title="Contextual Image Classification",
+    description="Upload an image, and the model will classify it based on the context."
 )
 
 # Launch the interface