Face embedding and comparison in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a face embedding model using a common deep learning library.

Computer Vision

model = [1]('facenet_keras.h5')

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

Aload_model

Bfit

Ccompile

Dpredict

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'fit' instead of 'load_model' which tries to train the model.

Using 'predict' which runs the model but does not load it.

The load_model function loads a saved deep learning model from a file.

2fill in blank

medium

Complete the code to preprocess a face image before embedding extraction.

Computer Vision

preprocessed_face = face_image[1](size=(160, 160))

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

Arotate

Bnormalize

Cresize

Dcrop

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'normalize' which changes pixel values but not size.

Using 'crop' which cuts part of the image but may not resize.

Face embedding models usually require images resized to a fixed size, here 160x160 pixels.

3fill in blank

hard

Fix the error in the code to compute the Euclidean distance between two face embeddings.

Computer Vision

distance = np.linalg.norm(embedding1 [1] embedding2)

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

A+

B-

C*

D/

Attempts:

3 left

💡 Hint

Common Mistakes

Using '+' which sums vectors instead of finding difference.

Using '*' or '/' which do element-wise multiplication or division.

Euclidean distance is the norm of the difference between two vectors.

4fill in blank

hard

Fill both blanks to create a dictionary of face embeddings for a list of face images.

Computer Vision

embeddings = {face_id: model.[1]([2](face)) for face_id, face in faces.items()}

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

Apredict

Bresize

Cpreprocess_input

Dfit

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'fit' which trains the model instead of predicting.

Using 'resize' as a model method which is incorrect.

We use preprocess_input to prepare the face image, then predict to get the embedding.

5fill in blank

hard

Fill all three blanks to compare two face embeddings and decide if they match based on a threshold.

Computer Vision

distance = np.linalg.norm(embedding1 [1] embedding2)
match = distance [2] threshold
result = 'Match' if match else [3]

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

A-

B<=

C'No Match'

D+

Attempts:

3 left

💡 Hint

Common Mistakes

Using '+' instead of '-' for distance calculation.

Using '>' instead of '<=' for threshold comparison.

Returning a boolean instead of a string for result.

We subtract embeddings to get distance, check if distance is less or equal to threshold, else return 'No Match'.