Image datasets (CIFAR-10, ImageNet) in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding CIFAR-10 Dataset Characteristics

Which of the following statements about the CIFAR-10 dataset is correct?

A CIFAR-10 is primarily used for natural language processing tasks.

B CIFAR-10 images are grayscale and have a resolution of 224x224 pixels.

C CIFAR-10 has 100 classes with 600 images each, all in black and white.

D CIFAR-10 contains 60,000 color images of size 32x32 pixels divided into 10 classes.

💡 Hint

Think about the number of classes and image size commonly used in beginner image classification tasks.

CIFAR-10 consists of 60,000 color images sized 32x32 pixels, split into 10 classes. It is widely used for image classification benchmarks.

❓ Predict Output

intermediate

Output Shape of CIFAR-10 Training Data

Given the following Python code snippet loading CIFAR-10 using TensorFlow, what is the shape of x_train?

Computer Vision

import tensorflow as tf
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()
print(x_train.shape)

A (50000, 32, 32, 3)

B (60000, 28, 28, 1)

C (10000, 32, 32, 3)

D (50000, 224, 224, 3)

💡 Hint

Recall the number of training images and their dimensions in CIFAR-10.

The CIFAR-10 training set contains 50,000 images, each 32x32 pixels with 3 color channels (RGB), so the shape is (50000, 32, 32, 3).

❓ Model Choice

advanced

Choosing a Model for ImageNet Classification

You want to train a deep learning model on the ImageNet dataset, which has over 1 million images and 1000 classes. Which model architecture is most suitable for this task?

A A small fully connected neural network with 2 hidden layers.

B A recurrent neural network (RNN) designed for text sequences.

C A convolutional neural network (CNN) like ResNet-50 pretrained on ImageNet.

D A simple logistic regression model.

💡 Hint

Consider the dataset size, image complexity, and model type best suited for images.

ResNet-50 is a deep CNN architecture designed for large-scale image classification tasks like ImageNet. It captures complex features and is pretrained for better performance.

❓ Metrics

advanced

Evaluating Model Accuracy on CIFAR-10

After training a model on CIFAR-10, you get the following confusion matrix summary for 10 classes. Which metric best summarizes the overall model performance?

A Perplexity, because it measures uncertainty in classification.

B Accuracy, because it shows the proportion of correctly classified images out of all images.

C Mean Squared Error, because it measures the average squared difference between predicted and true labels.

D BLEU score, because it evaluates image classification quality.

💡 Hint

Think about which metric is commonly used for classification tasks with multiple classes.

Accuracy is the standard metric for classification tasks, showing the fraction of correct predictions over total predictions.

🔧 Debug

expert

Identifying the Cause of Slow Training on ImageNet

You are training a CNN on ImageNet but notice training is extremely slow. Which of the following is the most likely cause?

A Using a very small batch size causing inefficient GPU utilization.

B Using data augmentation techniques like random flips and crops.

C Using pretrained weights from ImageNet for initialization.

D Using mixed precision training to speed up computation.

💡 Hint

Consider how batch size affects hardware usage during training.

Small batch sizes can cause GPUs to be underutilized, leading to slower training. Larger batches improve parallelism and speed.