Identify features of image classification solutions

(True/False) Image classification solutions require a large amount of labeled data for training.

• True

To achieve high accuracy, image classification solutions typically need a significant amount of labeled data to train the model.

(True/False) Image classification is used to detect and pinpoint the location of objects within an image.

• False

Image classification is used to categorize the entire image into classes, whereas object detection is used to find the location of objects within an image.

(Multiple Select) Which of the following are common features of image classification solutions?

• A) Feature extraction
• B) Object localization
• C) Accuracy assessment
• D) Transfer learning

Answer: A, C, D

Feature extraction and accuracy assessment are integral parts of image classification. Transfer learning is also a common technique used where a pre-trained model is adapted to a new problem.

(Single Select) Which Azure service is specifically designed for custom image classification?

• A) Azure Cognitive Services Text Analytics
• B) Azure Computer Vision
• C) Azure Custom Vision
• D) Azure Machine Learning Service

Answer: C

Azure Custom Vision is a service specifically designed for building and deploying custom image classification models.

(True/False) In image classification, deep learning models always outperform traditional machine learning models.

• False

Deep learning models often provide superior performance for complex image classification tasks, but traditional machine learning models may suffice for simple tasks or when computational resources are limited.

(Single Select) What is a convolutional neural network (CNN)?

• A) A natural language processing model
• B) A type of deep neural network especially suited for image classification
• C) A reinforcement learning algorithm
• D) A decision tree-based model

Answer: B

Convolutional neural networks (CNNs) are a class of deep learning neural networks that are particularly effective for image classification tasks.

(True/False) Transfer learning can only be applied in cases where the new dataset is very large.

• False

Transfer learning is especially useful when the new dataset is too small to train a fully-fledged model from scratch, as it leverages learned features from a related task with abundant data.

(True/False) Pre-trained models can reduce the time and resources required to develop an image classification solution.

• True

Pre-trained models have already learned features from vast datasets, reducing the need for extensive training on the task at hand.

(Single Select) When using Azure Custom Vision, you need to manually update the model for improved accuracy each time new data is available.

• A) True
• B) False

Answer: B

Azure Custom Vision can automatically retrain and improve the model as new data becomes available, which can be set up without manual intervention.

(Multiple Select) What are typical pre-processing steps in an image classification workflow?

• A) Normalization
• B) Dimensionality reduction
• C) Tokenization of text
• D) Image augmentation

Answer: A, B, D

Normalization, dimensionality reduction, and image augmentation are common pre-processing steps in image classification to prepare and enhance data before training.

(True/False) Azure Computer Vision API can be customized to detect brand-specific logos in images without any additional training.

• False

While Azure Computer Vision API has general logo detection capabilities, detecting brand-specific logos typically requires training a custom model, as can be done with Azure Custom Vision.

(True/False) The primary metric to evaluate an image classification model’s performance is always its accuracy.

• False

While accuracy is a common metric, other metrics like precision, recall, F1 score, and the area under the ROC curve can also be important depending on the specific application and context.