LargeScaleIncrementalLearning Evaluation metric for CELEB dataset.

Hi, what is the evaluation metric used for Celeb dataset? ( I guess for imagenet it is top-5?)

Nov 05 '19 14:11 brjathu

@brjathu The evaluation metric for MS-Celeb-1M is top-1 accuracy.

Nov 06 '19 14:11 wuyuebupt

Hi, nice work! May I know in the experiment for MS-Celeb-1M, do you still have the list of the 10,000 randomly selected classes? Appreciate it if you could share it.

Sep 25 '20 03:09 szewinglau

@szewinglau I'm just very lucky to find the training/validation list for the MS-Celeb-1M. I put them in the google drive https://drive.google.com/file/d/1F4AOBtFzJOYFXCzPsfXQdRFCdCdhwpM_/view?usp=sharing . The class is part of the string in each line, e.g. "m.0ch32zh" is the class name in a line "./val/m.0ch32zh/164097_0.jpg 0". Hope this help.

Sep 29 '20 23:09 wuyuebupt

Great, thanks a lot! I could not access the Google Drive. Please grant me access to your file (request sent from [email protected]). Thank you again.

Oct 01 '20 06:10 szewinglau

@szewinglau Sorry about that. I fix the link. You should be able to download now. Let me know if still any issue.

Oct 02 '20 01:10 wuyuebupt

@szewinglau Sorry about that. I fix the link. You should be able to download now. Let me know if still any issue.

I got it! Thanks a lot for your help!

Oct 02 '20 02:10 szewinglau