CV project workflow in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Key Step Identification in CV Project Workflow

Which step is typically the first in a computer vision project workflow?

A Training the model

B Collecting and labeling the dataset

C Evaluating the model performance

D Deploying the model to production

💡 Hint

Think about what you need before training a model.

The first step is to collect and label the dataset. Without data, you cannot train or evaluate a model.

❓ Model Choice

intermediate

Choosing a Model Architecture for Image Classification

You want to classify images into 10 categories. Which model architecture is most suitable to start with?

A Convolutional Neural Network (CNN)

B Recurrent Neural Network (RNN)

C Linear Regression

D K-Means Clustering

💡 Hint

Consider which model type is designed to process images.

CNNs are designed to handle image data by capturing spatial features. RNNs are for sequences, linear regression is for numeric prediction, and K-Means is unsupervised clustering.

❓ Metrics

advanced

Evaluating Model Performance with Accuracy and Confusion Matrix

After training a CV model, you get the following confusion matrix for 3 classes:

[[50, 2, 3], [4, 45, 1], [2, 3, 48]]

What is the overall accuracy?

A 0.91

B 0.86

C 0.95

D 0.89

💡 Hint

Accuracy = (sum of diagonal) / (sum of all elements).

Sum diagonal = 50+45+48=143; total sum = 50+2+3+4+45+1+2+3+48=158; accuracy = 143/158 ≈ 0.91.

🔧 Debug

advanced

Identifying the Bug in Data Augmentation Code

What error will this code raise when applying data augmentation using PyTorch transforms?

import torchvision.transforms as T
from PIL import Image
transform = T.Compose([
    T.RandomHorizontalFlip(p=0.5),
    T.ToTensor(),
    T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
])

image = Image.open('image.jpg')
augmented = transform(image)

A AttributeError: 'Image' object has no attribute 'shape'

B RuntimeError: image mode not supported

C No error, code runs successfully

D TypeError: normalize expects 3 channels but got 1

💡 Hint

Check the number of channels in the image and the mean/std length.

The image is likely RGB (3 channels), so mean and std should have 3 values each to match the channels, avoiding errors in Normalize.

❓ Hyperparameter

expert

Effect of Learning Rate on Training Stability

You train a deep CNN for object detection. Which learning rate choice is most likely to cause unstable training with loss oscillations?

A 0.0001

B 0.001

C 0.0005

D 0.01

💡 Hint

Higher learning rates can cause unstable updates.

A learning rate of 0.01 is often too high for deep CNNs, causing large weight updates and oscillating loss.