app.py

import gradio as gr
import cv2
import numpy as np
from tensorflow.keras.models import load_model



import gradio as gr
import cv2
import numpy as np
from tensorflow.keras.models import load_model

class_folders = ["Ajloun Castle", "Jeresh", "Petra", "Roman amphitheater", "Umm_Qais", "Wadi_Rum"]

Inception_path='./Models/new_efficientnet.h5'
VGGNet_path='./Models/new_vgg_16.hdf5'
ResNet_path= './Models/new_resnet50.hdf5'
EfficientNet_path= './Models/efficientnet-saved-model-06-val_acc-0.98.hdf5'

Inception_model = load_model(Inception_path)
VGGNet_model = load_model(VGGNet_path)
ResNet_model = load_model(ResNet_path)
Efficient_model = load_model(EfficientNet_path)

history=[]

def classify_selected_model(image, model_name):
    try:
        img = cv2.imread(image)
        img = cv2.resize(img, (150, 150))
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        img = np.expand_dims(img, axis=0)

        models = {
            "ResNet": ResNet_model,
            "EfficientNet": Efficient_model,
            "VGGNet": VGGNet_model,
            "InceptionV3": Inception_model
        }

        all_predictions = {}
        for model_key, model in models.items():
            predictions = model.predict(img)
            class_index = np.argmax(predictions[0])
            class_name = class_folders[class_index]
            probability = round(max(predictions[0]) * 100, 2)
            all_predictions[model_key] = probability

        min_probability = min(all_predictions.values())

        if min_probability < 70:  # Adjust the threshold as needed
            result_str = "Can't detect this image. Please try another one."
           
            return None, None, result_str

        else:
            class_index = np.argmax(predictions[0])
            class_name = class_folders[class_index]
            max_probability = round(max(predictions[0]) * 100, 2)
            result_str = f"{model_name}: {class_name} {max_probability}%"
            history.append(result_str)
            return class_name, max_probability,None, result_str

        # best_model = max(all_predictions, key=all_predictions.get)
        # class_name = class_folders[np.argmax(models[best_model].predict(img)[0])]
        # probability = all_predictions[best_model]

        # result_str = f"{best_model}: {class_name} {probability}%"
        # history.append(result_str)

        # return class_name, probability, history[-3:-1]

    except Exception as e:
        error_message = f"Error: {str(e)}"
        return None, None, [error_message]

image_input = gr.Image(type='filepath', label="Input Image")
model_dropdown = gr.Dropdown(["ResNet", "EfficientNet", "VGGNet", "InceptionV3"], label="Select Model")
result_textbox = gr.Textbox("Results", label="Error Message")

iface = gr.Interface(
    fn=classify_selected_model,
    inputs=[image_input, model_dropdown],
    outputs=[
        gr.Label(value="Archaeological Site"),
        gr.Label(value="Probability"),
        result_textbox
    ],
    examples=[
        ["Sites/Ajloun.png"],
        ["Sites/Jeresh.png"],
        ["Sites/Umm Qais.png"],
        ["Sites/Petra.png"],
        ["Sites/Roman Amphitheater.png"],
        ["Sites/Wadi Rum.png"]
    ],
    live=True,
    title="Image Classification for Archaeological Sites Interface"
)

iface.launch(debug=True, share=True)