Face recognition concept in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

How does face recognition typically work?

Which step is not usually part of a face recognition system?

A Translate faces into different languages

B Detect faces in an image

C Extract features from detected faces

D Compare extracted features to known faces

💡 Hint

Think about what face recognition systems do with faces, not unrelated tasks.

Face recognition systems detect faces, extract features, and compare them. Translating faces into languages is unrelated.

❓ Model Choice

intermediate

Choosing a model for face recognition

Which model type is best suited for extracting meaningful features from face images?

A K-Means Clustering

B Recurrent Neural Network (RNN)

C Linear Regression

D Convolutional Neural Network (CNN)

💡 Hint

Consider which model type is good at processing images.

CNNs are designed to process images and extract spatial features, making them ideal for face recognition.

❓ Metrics

advanced

Evaluating face recognition accuracy

Which metric best measures how well a face recognition system correctly identifies known faces?

A BLEU Score

B Mean Squared Error

C Accuracy

D Silhouette Score

💡 Hint

Think about a metric that measures correct predictions in classification.

Accuracy measures the proportion of correct identifications, suitable for face recognition. MSE is for regression, BLEU for language, Silhouette for clustering.

🔧 Debug

advanced

Troubleshooting face recognition model training

After training a face recognition model, the training accuracy is very high but test accuracy is very low. What is the most likely cause?

A The training data has too many classes

B The model is overfitting the training data

C The test data is corrupted

D The model is underfitting the training data

💡 Hint

Think about what happens when a model learns training data too well but fails on new data.

Overfitting means the model memorizes training data but does not generalize well to new data, causing low test accuracy.

❓ Predict Output

expert

Output of face embedding extraction code

What is the output shape of the face embedding vector extracted by this code?

Computer Vision

import numpy as np
from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from tensorflow.keras.preprocessing import image

model = ResNet50(weights='imagenet', include_top=False, pooling='avg')
img_path = 'face.jpg'
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
embedding = model.predict(x)
print(embedding.shape)

A (1, 2048)

B (224, 224, 3)

C (1, 1000)

D (7, 7, 2048)

💡 Hint

Pooling='avg' means the output is a vector, not a feature map.

ResNet50 with include_top=False and pooling='avg' outputs a 2048-length vector per image, shape (1, 2048).