app.py

import gradio as gr
import tensorflow as tf
from tensorflow.keras.applications import EfficientNetV2L
from tensorflow.keras.applications.efficientnet_v2 import preprocess_input, decode_predictions
from tensorflow.keras.preprocessing.image import img_to_array
from PIL import Image
import numpy as np

# Load the stronger pre-trained model (EfficientNetV2L)
model = EfficientNetV2L(weights="imagenet")

def predict_image(image):
    """
    Process the uploaded image and return the top 3 predictions as a dictionary.
    """
    try:
        # Preprocess the image
        image = image.resize((480, 480))  # EfficientNetV2L expects 480x480 input
        image_array = img_to_array(image)
        image_array = preprocess_input(image_array)  # Normalize the image
        image_array = np.expand_dims(image_array, axis=0)  # Add batch dimension

        # Get predictions
        predictions = model.predict(image_array)
        decoded_predictions = decode_predictions(predictions, top=3)[0]

        # Format predictions as a dictionary (label -> confidence)
        results = {label: float(confidence) for _, label, confidence in decoded_predictions}
        return results

    except Exception as e:
        return {"Error": str(e)}

# Create the Gradio interface
interface = gr.Interface(
    fn=predict_image,
    inputs=gr.Image(type="pil"),  # Accepts an image input
    outputs=gr.Label(num_top_classes=3),  # Shows top 3 predictions with confidence
    title="EfficientNetV2L Image Classifier",
    description="Upload an image, and the model will predict what's in the image with higher accuracy."
)

# Launch the Gradio app
if __name__ == "__main__":
    interface.launch()