StevenBanama
StevenBanama
> i cannot load pretrained model in file nets/test.py there exist a bug from tensorflow loader. I have hot fixed it. U need to update your repos.
> Hello, does facal loss address data imbalance? if data is balance, is facal loss useless? In this case, loss in the original paper should be adopted? @rongyixiao Q1:due to...
focal loss
> _No description provided._ @volcanoxs can you provide your command or any other information.
Agree with you. I dont experiment all regularizer due to limitation of time. I just wanna to keep bulk_feat as steady distribution , So you can change it.
profile(age and gender) Inference cost less than 20ms in arm-rk3399. MTCNN Detection cost much more time than profile model.
I have had test it in our inner dataset, which worked normally.But now I have changed my job and no GPU at hand.Maybe you can retrain or finetune it.
you can download dataset from pretrained model
@zhangyunming You should implement your own processer for custom dataset . Details can be found. in "https://github.com/StevenBanama/C3AE/blob/master/preproccessing/dataset_proc.py"
remove circus_zoos~~.U can find this module in dir named "detect" in this repo.