import gradio as gr import tensorflow as tf import numpy as np from PIL import Image model_path = "Bird_transfer_learning_NASNetLarge.keras" model = tf.keras.models.load_model(model_path) # Define the core prediction function def predict_bird(image): # Preprocess image print(type(image)) image = Image.fromarray(image.astype('uint8')) # Convert numpy array to PIL image image = image.resize((150, 150)) #resize the image to 28x28 and converts it to gray scale image = np.array(image) image = np.expand_dims(image, axis=0) # same as image[None, ...] # Predict prediction = model.predict(image) # No need to apply sigmoid, as the output layer already uses softmax # Convert the probabilities to rounded values prediction = np.round(prediction, 3) # Separate the probabilities for each class p_azure_jay = prediction[0][0] # Probability for class 'articuno' p_bald_eagle = prediction[0][1] # Probability for class 'moltres' p_jandaya_parakeet = prediction[0][2] # Probability for class 'zapdos' p_shoebill = prediction[0][3] # Probability for class 'zapdos' return {'Azure Jay': p_azure_jay, 'Bald Eagle': p_bald_eagle, 'Jandaya Parakeet': p_jandaya_parakeet, 'Shoebill': p_shoebill} # Create the Gradio interface input_image = gr.Image() iface = gr.Interface( fn=predict_bird, inputs=input_image, outputs=gr.Label(), examples=["images/Azure_jay_1.jpg", "images/Azure_jay_2.jpg", "images/Azure_jay_3.jpg", "images/Azure_jay_4.jpg", "images/Azure_jay_5.jpg", "images/Bald_eagle_1.jpg", "images/Bald_eagle_2.jpg", "images/Bald_eagle_3.jpg", "images/Bald_eagle_4.jpg", "images/Bald_eagle_5.jpg", "images/Jandaya_parakeet_1.jpg", "images/Jandaya_parakeet_2.jpg", "images/Jandaya_parakeet_3.jpg", "images/Jandaya_parakeet_4.jpg", "images/Jandaya_parakeet_5.jpg", "images/Shoebill_1.jpg", "images/Shoebill_2.jpg", "images/Shoebill_3.jpg", "images/Shoebill_4.jpg", "images/Shoebill_4.jpg"], description="Welcome to the Bird classifier. Upload an image of an Azure Jay, Bald Eagle, Jandaya Parakeet or a Shoebill to classify!") iface.launch()