A short quiz on vision system evaluation

Adrian F. Clark

1. What is 'ground truth'?

   • values obtained by an algorithm that are known to be true
   • images of the ground

   • data known to be correct

   • the true values obtained by an algorithm

   Ground truth are data (usually images) for which the correct answer is known; they are used for training and testing algorithms.

2. Which test is most appropriate for comparing algorithms' performances?

   • Canny's test

   • McNemar's test

   • Gauss's test
   • Laplace's test

   McNemar's test is the most appopriate test for comparing algorithms: it is a chi-squared test with one degree of freedom for paired data.

3. Which corner of a ROC curve indicates the best performance?

   • lower right

   • upper left

   • upper right
   • lower left

   We want the smallest number of false positives for the largest number of true positives, so the best performance is the upper left corner of the plot.

4. When using McNemar's test, what do we do if we want to see whether algorithms' performances differ?

   • Look up the Z-score in one-tailed tables
   • Look up the Z-score in Normal distribution tables

   • Look up the Z-score in two-tailed tables

   • Look up the Z-score in binomial tables

   If algorithms' performances differ, it doesn't matter which is better than the other, so we use two-tailed tables.

5. You are developing a automatic passport system for use by immigration, where pictures of people are compared to those in their passports. Which of the following is the best approach to take?

   • maximize the number of true positives, even if the false positive rate is high

   • minimize the number of false positives

   • minimize the number of false negatives
   • maximize the number of true positives

   For this type of application, we need to keep the number of false positives as low as possible; otherwise, we would admit lots of people who don't look like the picture on their passports.

6. When using McNemar's test, what do we do if we want to see whether one algorithm's performance is better than another's?

   • Look up the Z-score in one-tailed tables

   • Look up the Z-score in two-tailed tables
   • Look up the Z-score in Normal distribution tables
   • Look up the Z-score in binomial tables

   If we want to know that one algorithm's performance is better than another, we must use one-tailed tables.

7. What assumption underlies a null hypothesis test?

   • that there is no performance difference between algorithms

   • that the algorithms return a null result
   • that the algorithms don't work
   • that the algorithms differ in performance

   The null hypothesis test assumes that there is no difference in performance between algorithms, then examines whether the statistics support that assumption or provide evidence that is is wrong.

8. When evaluating vision systems, it is normal to:

   • have different training and test sets

   • use the same training and test sets
   • train on all data but test on only the test set
   • train on the training set and test using both training and test sets

   We know that algorithms work better on the data they were trained on than on unseen data; hence, we use different training and test sets.

9. What is a false negative?

   • A false result from an algorithm
   • A negative result that is false
   • A false result from an algorithm that should be negative

   • A false result from an algorithm that should have succeeded

   A false negative arises when an algorithm reports failure (or crashes) when it should have succeeded.

10. What is a false positive?

    • A positive result that is false

    • A true result from an algorithm that is incorrect

    • A false result from an algorithm
    • A false result from an algorithm that should be correct

    A false positive arises when an algorithm reports success but has actually found an incorrect result.