app.py

import os
import gradio as gr
import numpy as np
from tensorflow.keras.preprocessing.image import img_to_array
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
from tensorflow import keras
from huggingface_hub import hf_hub_download
from PIL import Image

# Disable GPU (forces CPU usage)
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

# Define class labels
class_labels = {0: 'cocci', 1: 'bacilli', 2: 'spirilla'}

# Function to load and use the model
def load():
    # Step 1: Download the model from Hugging Face Model Hub
    model_path = hf_hub_download(repo_id="D963934/bact_model", filename="bct_model.keras")
    
    # Step 2: Load the model using Keras
    model = keras.models.load_model(model_path)
    
    # Return the model for inference
    return model

# Function for image preprocessing and prediction
def predict(image):
    if image is None:
        print("Error: Image not loaded. Check the file path.")
    else: 
        # Step 1: Preprocess image for the model
        image = image.resize((224, 224))  # Resize to the expected input size
        image = img_to_array(image)  # Convert the image to an array
        image = np.expand_dims(image, axis=0)  # Add batch dimension
        image = preprocess_input(image)  # Preprocess for MobileNetV2
    
        # Load the model (assuming it is already loaded globally)
        model = load()

        # Step 2: Make predictions
        predictions = model.predict(image)
        predicted_class_idx = predictions.argmax()  # Class index with the highest probability
        confidence_score = predictions[0][predicted_class_idx]  # Confidence score
        #predicted_class = np.argmax(predictions, axis=1)[0]
        
        # Get class label
        predicted_class = class_labels[predicted_class_idx]
        #predicted_label = class_labels.get(predicted_class, 'Unknown')

        #return predicted_Class
        return f"Predicted Class: {predicted_class} (Confidence: {confidence_score:.2f})"

# Step 3: Define examples for Gradio

examples = [
    ["https://huggingface.co/spaces/D963934/BactSpace/resolve/main/img%203001.jpg"],
    ["https://huggingface.co/spaces/D963934/BactSpace/resolve/main/img%205000.jpg"],
    ["https://huggingface.co/spaces/D963934/BactSpace/resolve/main/img%20137.jpg"],
    ["https://huggingface.co/spaces/D963934/BactSpace/resolve/main/img%20352.jpg"],
    ["https://huggingface.co/spaces/D963934/BactSpace/resolve/main/img%20458.jpg"],
]
# Define the Gradio interface
iface = gr.Interface(fn=predict, 
                     inputs=gr.Image(type="pil"), 
                     outputs=gr.Text(label="Prediction Results:"),
                     title="Bacterial Identification",
                     description="Upload an image of bacteria to identify - cocci, bacilli, or spirilla",
                     examples=examples,
                     allow_flagging="never",
                     live=True)

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