lordfoogthe2st
/

dog-breed-identification

Model card Files Files and versions Community

dog-breed-identification / app.py

lordfoogthe2st's picture

create app.py

d4da8ae 11 months ago

raw history blame contribute delete

1.65 kB

	def complete_model_pipeline(image):
	"""
	Complete model pipeline for deploying the model using Gradio.

	This function takes an image as input, saves it temporarily to a file,
	loads the image as a test sample, performs prediction using the model,
	and then removes the temporarily saved image. The function returns the
	predicted dog breed label.

	Args:
	image (np.ndarray): The NumPy array representing the input image.

	Returns:
	str: The predicted dog breed label.

	Parameters:
	- image: A NumPy array representing the input image.

	Note: This function is designed to be used with Gradio to deploy the model
	and obtain predictions interactively.

	"""

	# Define the file path to temporarily save the image
	image_file_path = '/content/drive/MyDrive/Colab Notebooks/dog-vision-project/custom_images/myImage.jpg'

	# Convert the input NumPy array to a Pillow image object and save it as a file
	image = Image.fromarray(image)
	image.save(image_file_path)

	# Prepare test data for the model using the saved image file
	test_data = get_data_batches(X=[image_file_path], test_data=True, batch_size=1, deployment = True)

	# Perform prediction using the model
	predictions = model.predict(test_data, verbose = 0)

	# Get the index of the highest predicted value and obtain the corresponding breed label
	predicted_label_index = np.argmax(predictions[0])
	predicted_breed_label = unique_breeds[predicted_label_index]

	# Remove the temporarily saved image file
	os.remove(image_file_path)

	return format_underscores(predicted_breed_label)