import gradio as gr
import tensorflow as tf
import numpy as np
import pickle
import cv2

# Load model and class names
model = tf.keras.models.load_model("model.h5")

with open("class_names.pkl", "rb") as f:
    class_names = pickle.load(f)

# Image preprocessing function
def preprocess_image(img):
    img = cv2.resize(img, (224, 224))
    img = img / 255.0
    return np.expand_dims(img, axis=0)

# Prediction function
def predict(img):
    img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)  # Convert PIL image to OpenCV format
    processed = preprocess_image(img)
    prediction = model.predict(processed)[0]
    predicted_label = class_names[np.argmax(prediction)]
    confidence = float(np.max(prediction)) * 100
    return f"Predicted: {predicted_label} ({confidence:.2f}%)"

# Gradio interface
interface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="numpy", label="Upload an Animal Image"),
    outputs="text",
    title="Animal Classifier with ResNet50",
    description="Upload an image of an animal to classify using a pretrained ResNet50 model."
)

interface.launch()