search

xiaomengyc / ACoL

(CVPR2018) Adversarial Complementary Learning for Weakly Supervised Object Localization
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_Adversarial_Complementary_Learning_CVPR_2018_paper.pdf
MIT License
164 stars 30 forks source link

Regarding training loss #4

Open laukun opened 5 years ago

laukun commented 5 years ago

Thank author for the code. From the figures showing CAM and yours, there's not much difference on the mined extent of the object. Do you have the loss curves of training on some dataset before and after adding ACoL?

xiaomengyc commented 5 years ago

@laukun Sorry, I did not compare the loss curves between the two methods. Although these two methods have the same mathematical formalization, the network structures are not exactly the same. Also, the training process will be influenced by the stochastic optimization process. So, I think there will be some differences between the two convergent curves.