xiaobaishu0097 / megc2019 Goto Github PK

Could you share me with the code of the apex frame magnification?

There seems to be an issue with data leakage here. In line 406 and 407 the best f1-score is updated inside the LOSO loop. Essentially, this is early stopping where the hyperparameter for stopping epoch value is found using the testing data. I rerun the experiment without these lines and achieved a 0.59 f1-score for MEGC.

To use early stopping, one would have to use separate validation data that finds the stopping epoch value, which could then be used for the testing data as well.

Hi ,thank you for your work.could you give a link for SAMM and SMIC datasets links? thank you for your reply