Computer Vision Quiz

Question 1A camera supplies an image 256 pixels wide and 192 pixels high, with 3 channel values per pixel. How many channel values does it contain? Enter the count rounded to the nearest whole number.

Enter a number in values.

Answers within 0.5 values of the expected value are accepted.

Question 2A program must outline the exact painted region on a panel, following its irregular edge. Which output fits that task?

An image-wide average of the recorded brightnessA rectangle enclosing the panel’s painted patchA category score for the panel as a wholeAn image mask identifying the painted area

Question 3During model fitting, predicted labels disagree with the recorded labels. What does the optimizer adjust?

Pixel intensities in the stored training imagesThe recorded answers to match the current predictionsThe weights and biases that control the model’s outputThe definitions of the classes used for evaluation

Question 4During inspection, a trained network produces a strong response at an edge. What does this activation represent?

A physical depth reading encoded in the edge itselfA learned feature responding to a numerical patternA miniature object drawing held inside the activated unitAn object name supplied directly by the camera

Question 5An inspection system succeeds at its training station but fails under a replacement lamp. What is the strongest next test?

Evaluation under the new illumination with known outcomesApproval based on the model file remaining unchangedComparison of the two lamps’ purchase pricesA longer run of the original station’s saved examples

Question 6A person reaches around a partly hidden cup, while software labels a fixed crop as a cup. What follows from their matching labels?

Both identify the cup in the limited task demonstratedThe software and observer share experience of handling cupsBoth have checked whether the cup’s support will stay stableBoth systems have inferred how the unseen handle supports a grip

Question 7Several almost identical frames from one recording are divided between fitting and evaluation. Why is the resulting score questionable?

A larger number of frames guarantees independent visual evidenceAdjacent frames test transfer to unfamiliar scenesFamiliar scene details inflate the apparent generalizationThe evaluation has become harder by using familiar backgrounds

Question 8Searching enrolled faces returns the nearest enrolled template. How should that result be interpreted?

A legal permission granted by the similarity rankingA detected face location establishing the person’s nameA candidate match needing checks suited to the decisionA confirmed identity established by being nearest in the database

Question 9During operation, a camera image becomes badly blurred. Which response respects the system’s limits?

Increase the confidence attached to the previous predictionInterpret the blur as confirmation that the target is absentAssign the most frequent label to certify the current sceneDefer the decision when the visible evidence is inadequate

Question 10A vision score is described as a probability of being correct. What evidence supports that description?

The size of the parameter file stored on diskObserved correctness matching the assigned confidence bandsFaster execution when producing the requested answerAdditional decimal places in the displayed output

Answer every question to see and share your score.

Computer Vision Quiz

How does it work?

What will you get?