This folder contains the following 10 comma separated txt files which can be used to perform the evaluation in Tables 2 and 3 of the paper[1] : 

veriflist_test_random_seenheard.txt 
veriflist_test_random_unseenunheard.txt

veriflist_test_G_seenheard.txt
veriflist_test_G_unseenunheard.txt

veriflist_test_N_seenheard.txt
veriflist_test_N_unseenunheard.txt

veriflist_test_A_seenheard.txt
veriflist_test_A_unseenunheard.txt

veriflist_test_GNA_seenheard.txt
veriflist_test_GNA_unseenunheard.txt


The naming convention of the files is veriflist_test_{DEMOGRAPHICCRITERION}_{TESTSET}.txt

There are 5 different demographic criteria as outlined in Table 3 of the paper[1]: 
(1)random (2)G (3)N (4)A (5)GNA
There are 2 different test sets: 
(1) seenheard (2) unseenunheard 

The format of the files is: 
id,voice-speaker-id,face-speaker-id,label,voice-path,face-path

Please cite [1] below if you use this evaluation split:

[1] Nagrani, Arsha, Samuel Albanie, and Andrew Zisserman. "Learnable PINs: Cross-Modal Embeddings for Person Identity." ECCV, 2018.