Model evaluation best practices in Computer Vision - Cheat Sheet & Quick Revision

Recall & Review

beginner

Why is it important to split your dataset into training, validation, and test sets?

Splitting the dataset helps to train the model on one part (training), tune parameters on another (validation), and finally evaluate performance on unseen data (test) to avoid overfitting and get a realistic measure of how the model will perform in real life.

beginner

What does 'overfitting' mean in model evaluation?

Overfitting happens when a model learns the training data too well, including noise and details that don't generalize. This causes poor performance on new, unseen data.

intermediate

What is the purpose of using metrics like accuracy, precision, recall, and F1-score in computer vision?

These metrics help measure how well the model predicts. Accuracy shows overall correctness, precision measures how many predicted positives are true, recall shows how many actual positives were found, and F1-score balances precision and recall.

intermediate

Why should you use cross-validation in model evaluation?

Cross-validation splits data into multiple parts and trains/tests the model several times. This gives a better estimate of model performance by reducing bias from a single train-test split.

beginner

What is the difference between validation data and test data?

Validation data is used during model training to tune parameters and make decisions. Test data is kept separate and used only once at the end to evaluate the final model's performance.

What is the main goal of splitting data into training and test sets?

A To check how well the model performs on unseen data

B To make the training faster

C To increase the size of the dataset

D To reduce the number of features

Splitting data allows us to test the model on data it hasn't seen before, which shows how well it will perform in real situations.

Which metric balances precision and recall in classification tasks?

A F1-score

B Accuracy

C Loss

D Mean Squared Error

F1-score combines precision and recall into a single metric to balance false positives and false negatives.

What does overfitting cause in a model?

A Better performance on new data

B Poor performance on new, unseen data

C Poor performance on training data

D Faster training

Overfitting means the model fits training data too closely and fails to generalize to new data.

Why is cross-validation useful?

A It increases dataset size

B It removes irrelevant features

C It speeds up training

D It reduces bias in performance estimates

Cross-validation tests the model multiple times on different data splits to give a more reliable performance estimate.

When should you use the test dataset?

A To increase training speed

B During model training to adjust parameters

C To evaluate the final model after training

D To create new features

Test data is used only once after training to check how well the model performs on unseen data.

Explain why splitting data into training, validation, and test sets is important in model evaluation.

Think about how each set helps the model learn and be tested fairly.

Describe the difference between precision and recall and why both are important in evaluating a computer vision model.

Consider how mistakes in predictions affect model usefulness.