QCM : Advanced Image Recognition and Classification — 10 questions

Questions et réponses du QCM

1. What is an image matching technique primarily concerned with?

Matching algorithms comparing features in images
Classifying images into categories
Interest point detection in images
Feature descriptor computation for images

Interest point detection in images

Explication

Interest point detection is the core process in many image matching techniques, involving identifying salient points in images that can be reliably matched across different images. This step is fundamental before feature description and matching.

2. Who developed the Scale-Invariant Feature Transform (SIFT) as a feature descriptor?

Lowe (2004)
Alan Turing
Charles Darwin
Kurt Gödel

Lowe (2004)

Explication

Lowe (2004) is the author who developed the SIFT feature descriptor, which is designed to be invariant to scale and rotation. The other options are notable figures in different fields and are not related to the development of SIFT.

3. What is the primary role of image classification in computer vision?

To assign a label or category to an entire image
To detect and localize objects within an image
To segment an image into different regions
To extract features from images for matching

To assign a label or category to an entire image

Explication

The main purpose of image classification is to assign a label or category to the entire image based on its visual content, which helps in organizing and understanding images at a high level.

4. Which learning paradigm was established first, followed by the others in chronological order?

Unsupervised learning, then supervised learning, then semi-supervised learning
Supervised learning, then unsupervised learning, then semi-supervised learning
Semi-supervised learning, then unsupervised learning, then supervised learning
Semi-supervised learning, then supervised learning, then unsupervised learning

Supervised learning, then unsupervised learning, then semi-supervised learning

Explication

Supervised learning was established first, with foundational work in the mid-20th century on training models with labeled data. Unsupervised learning came later, focusing on discovering structure in unlabeled data. Semi-supervised learning is the most recent, developed to leverage both labeled and unlabeled data, making the chronological order: supervised, then unsupervised, then semi-supervised.

5. How does the decision boundary of a general linear classifier differ from the maximum margin hyperplane used in support vector machines?

Linear classifiers do not have a decision boundary, but SVMs do.
The decision boundary in a linear classifier is always a hyperplane, while in SVMs it is a nonlinear curve.
The maximum margin hyperplane in SVMs is specifically chosen to maximize the distance to support vectors, whereas a general linear classifier's boundary does not necessarily maximize this margin.
The decision boundary of a linear classifier is always curved, while in SVMs it is always a straight line.

The maximum margin hyperplane in SVMs is specifically chosen to maximize the distance to support vectors, whereas a general linear classifier's boundary does not necessarily maximize this margin.

Explication

The maximum margin hyperplane in SVMs is explicitly chosen to maximize the distance to the nearest data points (support vectors), which is a defining characteristic of SVMs. In contrast, a general linear classifier's decision boundary is any hyperplane that separates classes but does not necessarily maximize this margin.

6. Who is credited with proposing support vector machines?

Vladimir Vapnik and Alexey Chervonenkis
Vladimir Vapnik and Corinna Cortes
Vladimir Vapnik and Bernhard Schölkopf
Vladimir Vapnik and Geoffrey Hinton

Vladimir Vapnik and Corinna Cortes

Explication

Support vector machines were proposed by Vladimir Vapnik and Corinna Cortes in 1995, establishing the framework for maximum margin classifiers. The other options include notable researchers but are not credited with proposing SVMs.

7. What is a primary effect of using ensemble methods in image classification tasks?

They eliminate the need for feature extraction in classification.
They improve overall accuracy and robustness of classifiers.
They simplify the model architecture by using fewer classifiers.
They significantly reduce training time for models.

They improve overall accuracy and robustness of classifiers.

Explication

Ensemble methods such as boosting and cascading classifiers are designed to improve the accuracy and robustness of classifiers by combining multiple models, which reduces overfitting and increases generalization performance.

8. Which of the following best describes how to apply the Viola-Jones method in practice for object detection?

Apply convolutional neural networks trained on large datasets to classify entire images without localization.
Use Histogram of Oriented Gradients (HOG) features with a linear SVM for object detection.
Use Haar features with a cascade of classifiers trained with AdaBoost to detect objects like faces in real-time.
Employ a sliding window approach with deep learning features extracted from a pre-trained CNN.

Use Haar features with a cascade of classifiers trained with AdaBoost to detect objects like faces in real-time.

Explication

The Viola-Jones method applies Haar features combined with a cascade of classifiers trained with AdaBoost for rapid, real-time detection of objects such as faces. It is a practical, feature-based detection framework that is widely used for real-time applications.

9. Which of the following best describes the key components used to evaluate the properties of performance metrics in classification tasks?

True Positives, True Negatives, False Positives, False Negatives
Accuracy, Precision, Recall, F1 Score
Model parameters, Hyper-parameters, Learning rate, Number of epochs
Training data, Validation data, Test data, Cross-validation results

True Positives, True Negatives, False Positives, False Negatives

Explication

The key components used to evaluate the properties of performance metrics are True Positives, True Negatives, False Positives, and False Negatives, as they form the confusion matrix from which metrics like sensitivity and specificity are derived. These components directly measure the classifier's performance in different aspects, making them fundamental to understanding and calculating performance metrics.

10. What is the kernel trick in support vector machines (SVM)?

A technique for implicitly mapping data into a higher-dimensional space to enable linear separation
A method for explicitly transforming data into a higher-dimensional space
A function used to measure similarity between data points in the original space
A way to compute the distance between data points in the original space

A technique for implicitly mapping data into a higher-dimensional space to enable linear separation

Explication

The kernel trick in support vector machines is a technique that allows the computation of inner products in a high-dimensional feature space without explicitly performing the transformation. This enables the SVM to find non-linear decision boundaries efficiently by implicitly mapping data into a higher-dimensional space where it becomes linearly separable.

Révisez avec les flashcards

Mémorisez les réponses avec 20 flashcards sur Advanced Image Recognition and Classification.

Interest point detection — purpose?

Identify repeatable, distinctive features in images.

Harris detector — key idea?

Detect corners via intensity autocorrelation analysis.

Scale-adapted Harris — extension?

Detects features across multiple scales.

Voir les flashcards →

Approfondir avec la fiche

Consultez la fiche de révision complète sur Advanced Image Recognition and Classification.

Voir la fiche →

Cours similaires

Introduction à l'Intelligence Artificielle Symbolique

Intelligence Artificielle

Introduction aux documents et sécurité en construction

Bases de données

Introduction à l'Informatique et IA

Introduction aux marchés publics et réglementations du bâtiment

Bases de données

Gestion des imprimantes Windows et PDF

Introduction aux bases de données relationnelles

Bases de données

Crée tes propres QCM

Importe ton cours et l'IA génère des QCM avec corrections en 30 secondes.

Générateur de QCM