Re-Identification Rank-1 Accuracy

Problem

Compute rank-1 accuracy for person re-identification.

Each row of preds is a ranked candidate list and gts[i] is the true identity of query i. Count queries whose top candidate matches the ground truth and return the fraction rounded to 4 decimal places. An empty input returns 0.0.

Hint

Only the first entry of each ranked list matters for rank-1.

Test cases

case 1

in: [[[1,2],[3,4]],[1,3]]

expected: 1

case 2

in: [[[2,1],[3,4]],[1,4]]

expected: 0

case 3

in: [[[1],[2],[3]],[1,2,3]]

expected: 1

case 4

in: [[[1,2],[2,1],[3,4]],[2,1,3]]

expected: 0.3333

Starter codePython

def rank1_accuracy(preds, gts):
    # Your code here
    pass

4 test cases · run your code against them in the editor