Closed tymanman closed 2 years ago
Hey sorry for the delayed response. Would you like to provide us a bit more details about your training procedure and configs/logs for a closer inspection?
Thanks!
There was a mistake I made when loading the pretrained model.Thank u.
发自我的iPhone
------------------ Original ------------------ From: Xin Wang @.> Date: Fri,May 14,2021 10:34 AM To: ucbdrive/few-shot-object-detection @.> Cc: tymanman @.>, Author @.> Subject: Re: [ucbdrive/few-shot-object-detection] Get zero ap for all classes after finetuning stage (#111)
Hey sorry for the delayed response. Would you like to provide us a bit more details about your training procedure and configs/logs for a closer inspection?
Thanks!
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
I follow the guidance to pretrain a base model for 15 base classes and convert the model by random initialization, finally I finetune the model on all classes.However the ap of all classes is zero, neither the auto evaluation at the end of training nor a repeated evaluation by the script of tool_test.How come?