Mobile deployment (TFLite, Core ML) in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to convert a TensorFlow model to TensorFlow Lite format.

Computer Vision

import tensorflow as tf

model = tf.keras.models.load_model('model.h5')
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.[1]()

Drag options to blanks, or click blank then click option'

Acompile

Bconvert

Csave

Dfit

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'compile()' instead of 'convert()' which is for preparing the model for training.

Using 'save()' which saves the model but does not convert it.

Using 'fit()' which trains the model.

The method 'convert()' converts the Keras model to TensorFlow Lite format.

2fill in blank

medium

Complete the code to load a Core ML model in Swift for iOS deployment.

Computer Vision

import CoreML

guard let model = try? [1](configuration: MLModelConfiguration()) else {
    fatalError("Failed to load model")
}

Drag options to blanks, or click blank then click option'

AMyModel

BMLModel

CVNCoreMLModel

DMLModelConfiguration

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'MLModel' directly instead of the generated model class.

Confusing 'VNCoreMLModel' which is for Vision framework.

Using 'MLModelConfiguration' as a model instance.

You instantiate your specific Core ML model class (e.g., 'MyModel') to load it.

3fill in blank

hard

Fix the error in the TensorFlow Lite interpreter initialization code.

Computer Vision

import tensorflow as tf

interpreter = tf.lite.Interpreter(model_path='model.tflite')
interpreter.[1]()

Drag options to blanks, or click blank then click option'

Aprepare

Bload_tensors

Cinitialize

Dallocate_tensors

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'load_tensors()' which does not exist.

Using 'initialize()' or 'prepare()' which are not valid methods.

You must call 'allocate_tensors()' to prepare the interpreter before inference.

4fill in blank

hard

Fill both blanks to create a dictionary mapping input names to input data for TFLite inference.

Computer Vision

input_details = interpreter.get_input_details()
input_data = [1]
inputs = {input_details[0]['name']: [2]

Drag options to blanks, or click blank then click option'

Anp.array([1.0, 2.0, 3.0], dtype=np.float32)

Binput_data

Cnp.array([0.0, 0.0, 0.0], dtype=np.float32)

Dinput_details

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'input_details' as input data.

Using a zero array instead of actual input values.

Not matching the input name to the data variable.

First, input_data is created as a numpy array, then the dictionary maps the input name to input_data.

5fill in blank

hard

Fill all three blanks to run inference and get the output tensor from a TFLite interpreter.

Computer Vision

interpreter.set_tensor(input_details[0]['index'], [1])
interpreter.[2]()
output_data = interpreter.get_tensor(output_details[0]['[3]'])

Drag options to blanks, or click blank then click option'

Ainput_data

Binvoke

Cindex

Doutput

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'output' instead of 'index' to get tensor.

Calling a wrong method instead of 'invoke()'.

Passing wrong variable to set_tensor.

You set the input tensor with input_data, call 'invoke()' to run inference, then get the output tensor by 'index'.