SQUEEZENET1.1

import cv2

import numpy as np

from openvino.runtime import Core, Tensor

# Function to preprocess the image

def preprocess_image(image_path):

    # Load the image

    image = cv2.imread(image_path)

    # Resize the image to 227x227 (input size for SqueezeNet)

    image_resized = cv2.resize(image, (227, 227))

    # Convert from BGR to RGB

    image_rgb = image_resized #cv2.cvtColor(image_resized, cv2.COLOR_BGR2RGB)

    # Normalize the image

    image_normalized = image_rgb.astype(np.float32)  

    # Change shape to (1, 3, 227, 227)

    image_input = np.transpose(image_normalized, (2, 0, 1)) # Shape becomes (3, 227, 227)

    image_input = np.expand_dims(image_input, axis=0) # Add batch dimension

    return image_input

# Load the OpenVINO runtime

ie = Core()

# Load the model

model_xml = 'models/squeezenet1.1/FP16/squeezenet1.1.xml'

model_bin = 'models/squeezenet1.1/FP16/squeezenet1.1.bin'

model = ie.read_model(model=model_xml, weights=model_bin)

# Compile the model for the target device (e.g., CPU)

compiled_model = ie.compile_model(model, device_name='CPU')

# Create an inference request

infer_request = compiled_model.create_infer_request()

# Preprocess the input image

image_input = preprocess_image("[Dataset]_Module22_images/cat.jpeg")

# Create an OpenVINO tensor from the NumPy array

input_tensor = Tensor(image_input)

# Set the input tensor

infer_request.set_input_tensor(input_tensor)

# Perform inference

infer_request.infer()

# Get the output tensor

output_tensor = infer_request.get_output_tensor(0) # Assuming single output

output_data = output_tensor.data

# Post-process the output

predictions = np.argmax(output_data, axis=1)

print("Predicted class:", predictions)

# Load the labels

def load_labels(labels_file):

    with open(labels_file, 'r') as f:

        labels = [line.strip() for line in f.readlines()]

    return labels

predictions.resize(1)

 

# Load labels and get the predicted label

labels = load_labels('models/squeezenet1.1/FP32/squeezenet1.1.labels') # Replace with the correct path to your labels file

predicted_label = labels[predictions[0]] # Assuming single image input

print("Predicted class:", predicted_label)