nesl / asvspoof2019

Our submission to the ASVspoof 2019: Automatic Speaker Verification Spoofing and Countermeasures Challenge
96 stars 40 forks source link

how to use the evaluation set #17

Open 1154015738 opened 4 years ago

1154015738 commented 4 years ago

After i run the code in the dev, i want to test it in the eval,but the lfcc_result is like: PA_0016 8.686298370361328 PA_0029 -211.11810302734375 PA_0044 -44.42363739013672

and when i run it on the dev,the result is like: PA_D_0000003 - bonafide 29.773143768310547 PA_D_0000004 - bonafide 15.644768714904785 it can work

because the lfcc_result on the eval lost two lists ,so i can't run the evaluate code. by check the code, i found those code make the lfcc_result

with open(save_path, 'w') as fh: for f, s, k, cm in zip(fname_list, sys_id_list, key_list, score_list): if not dataset.is_eval: fh.write('{} {} {} {}\n'.format(f, s, k, cm)) else: fh.write('{} {}\n'.format(f, cm)) print('Result saved to {}'.format(save_path))` so i want to ask why not use s and k lists in the evaluate set,and how to modify the code to let it work.