app.py

import gradio as gr
import tensorflow as tf
import numpy as np
from PIL import Image

model_path = "transferlearning_tools.keras"
model = tf.keras.models.load_model(model_path)
 
# Define the core prediction function
def predict_tools(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 150x150
    image = np.array(image)
    image = np.expand_dims(image, axis=0)  # Expand dimensions to match the model input shape
    
    # Predict
    prediction = model.predict(image)
    
    # Print the shape of the prediction to debug
    print(f"Prediction shape: {prediction.shape}")
    
    # Assuming the output is already softmax probabilities
    probabilities = prediction[0]
    
    # Print the probabilities array to debug
    print(f"Probabilities: {probabilities}")
    
    # Assuming your model was trained with these class names
    class_names = ['Gasoline can', 'Hammer', 'Rope', 'Screw driver', 'Wrench'] 
    
    # Create a dictionary of class probabilities
    result = {class_names[i]: float(probabilities[i]) for i in range(len(class_names))}
    
    return result
 

# Create the Gradio interface
input_image = gr.Image()
iface = gr.Interface(
    fn=predict_tools,
    inputs=input_image, 
    outputs=gr.Label(),
    examples=["tools_examples/Gasoline can 1.jpg", 
              "tools_examples/Gasoline can 2.jpg", 
              "tools_examples/Hammer 1.jpg", 
              "tools_examples/Hammer 2.jpg", 
              "tools_examples/Rope 1.jpg", 
              "tools_examples/Rope 2.jpg", 
              "tools_examples/Screw driver 1.jpg", 
              "tools_examples/Screw driver 2.jpg", 
              "tools_examples/Wrench 1.jpg", 
              "tools_examples/Wrench 2.jpg"],   
    description="A simple mlp classification model for image classification using the mnist dataset.")
iface.launch()